Index: projects/clang-trunk/Makefile.inc1 =================================================================== --- projects/clang-trunk/Makefile.inc1 (revision 283870) +++ projects/clang-trunk/Makefile.inc1 (revision 283871) @@ -1,2226 +1,2237 @@ # # $FreeBSD$ # # Make command line options: # -DNO_CLEANDIR run ${MAKE} clean, instead of ${MAKE} cleandir # -DNO_CLEAN do not clean at all # -DDB_FROM_SRC use the user/group databases in src/etc instead of # the system database when installing. # -DNO_SHARE do not go into share subdir # -DKERNFAST define NO_KERNEL{CONFIG,CLEAN,DEPEND,OBJ} # -DNO_KERNELCONFIG do not run config in ${MAKE} buildkernel # -DNO_KERNELCLEAN do not run ${MAKE} clean in ${MAKE} buildkernel # -DNO_KERNELDEPEND do not run ${MAKE} depend in ${MAKE} buildkernel # -DNO_KERNELOBJ do not run ${MAKE} obj in ${MAKE} buildkernel # -DNO_PORTSUPDATE do not update ports in ${MAKE} update # -DNO_ROOT install without using root privilege # -DNO_DOCUPDATE do not update doc in ${MAKE} update # -DWITHOUT_CTF do not run the DTrace CTF conversion tools on built objects # LOCAL_DIRS="list of dirs" to add additional dirs to the SUBDIR list # LOCAL_ITOOLS="list of tools" to add additional tools to the ITOOLS list # LOCAL_LIB_DIRS="list of dirs" to add additional dirs to libraries target # LOCAL_MTREE="list of mtree files" to process to allow local directories # to be created before files are installed # LOCAL_TOOL_DIRS="list of dirs" to add additional dirs to the build-tools # list # METALOG="path to metadata log" to write permission and ownership # when NO_ROOT is set. (default: ${DESTDIR}/METALOG) # TARGET="machine" to crossbuild world for a different machine type # TARGET_ARCH= may be required when a TARGET supports multiple endians # BUILDENV_SHELL= shell to launch for the buildenv target (def:/bin/sh) # WORLD_FLAGS= additional flags to pass to make(1) during buildworld # KERNEL_FLAGS= additional flags to pass to make(1) during buildkernel # # The intended user-driven targets are: # buildworld - rebuild *everything*, including glue to help do upgrades # installworld- install everything built by "buildworld" # doxygen - build API documentation of the kernel # update - convenient way to update your source tree (eg: svn/svnup) # # Standard targets (not defined here) are documented in the makefiles in # /usr/share/mk. These include: # obj depend all install clean cleandepend cleanobj .if !defined(TARGET) || !defined(TARGET_ARCH) .error "Both TARGET and TARGET_ARCH must be defined." .endif .include "share/mk/src.opts.mk" .include .include # We must do lib/ and libexec/ before bin/, because if installworld # installs a new /bin/sh, the 'make' command will *immediately* # use that new version. And the new (dynamically-linked) /bin/sh # will expect to find appropriate libraries in /lib and /libexec. # SRCDIR?= ${.CURDIR} .if defined(SUBDIR_OVERRIDE) SUBDIR= ${SUBDIR_OVERRIDE} .else SUBDIR= lib libexec SUBDIR+=bin .if ${MK_GAMES} != "no" SUBDIR+=games .endif .if ${MK_CDDL} != "no" SUBDIR+=cddl .endif SUBDIR+=gnu include .if ${MK_KERBEROS} != "no" SUBDIR+=kerberos5 .endif .if ${MK_RESCUE} != "no" SUBDIR+=rescue .endif SUBDIR+=sbin .if ${MK_CRYPT} != "no" SUBDIR+=secure .endif .if !defined(NO_SHARE) SUBDIR+=share .endif SUBDIR+=sys usr.bin usr.sbin .if ${MK_TESTS} != "no" SUBDIR+= tests .endif .if ${MK_OFED} != "no" SUBDIR+=contrib/ofed .endif # # We must do etc/ last for install/distribute to work. # SUBDIR+=etc # Local directories are last, since it is nice to at least get the base # system rebuilt before you do them. .for _DIR in ${LOCAL_DIRS} .if exists(${.CURDIR}/${_DIR}/Makefile) SUBDIR+= ${_DIR} .endif .endfor # Add LOCAL_LIB_DIRS, but only if they will not be picked up as a SUBDIR # of a LOCAL_DIRS directory. This allows LOCAL_DIRS=foo and # LOCAL_LIB_DIRS=foo/lib to behave as expected. .for _DIR in ${LOCAL_DIRS:M*/} ${LOCAL_DIRS:N*/:S|$|/|} _REDUNDENT_LIB_DIRS+= ${LOCAL_LIB_DIRS:M${_DIR}*} .endfor .for _DIR in ${LOCAL_LIB_DIRS} .if empty(_REDUNDENT_LIB_DIRS:M${_DIR}) && exists(${.CURDIR}/${_DIR}/Makefile) SUBDIR+= ${_DIR} .else .warning ${_DIR} not added to SUBDIR list. See UPDATING 20141121. .endif .endfor .endif .if defined(NOCLEAN) NO_CLEAN= ${NOCLEAN} .endif .if defined(NO_CLEANDIR) CLEANDIR= clean cleandepend .else CLEANDIR= cleandir .endif LOCAL_TOOL_DIRS?= BUILDENV_SHELL?=/bin/sh SVN?= /usr/local/bin/svn SVNFLAGS?= -r HEAD MAKEOBJDIRPREFIX?= /usr/obj .if !defined(OSRELDATE) .if exists(/usr/include/osreldate.h) OSRELDATE!= awk '/^\#define[[:space:]]*__FreeBSD_version/ { print $$3 }' \ /usr/include/osreldate.h .else OSRELDATE= 0 .endif .endif .if !defined(VERSION) REVISION!= ${MAKE} -C ${SRCDIR}/release -V REVISION BRANCH!= ${MAKE} -C ${SRCDIR}/release -V BRANCH SRCRELDATE!= awk '/^\#define[[:space:]]*__FreeBSD_version/ { print $$3 }' \ ${SRCDIR}/sys/sys/param.h VERSION= FreeBSD ${REVISION}-${BRANCH:C/-p[0-9]+$//} ${TARGET_ARCH} ${SRCRELDATE} .endif KNOWN_ARCHES?= aarch64/arm64 amd64 arm armeb/arm armv6/arm armv6hf/arm i386 i386/pc98 mips mipsel/mips mips64el/mips mips64/mips mipsn32el/mips mipsn32/mips powerpc powerpc64/powerpc sparc64 .if ${TARGET} == ${TARGET_ARCH} _t= ${TARGET} .else _t= ${TARGET_ARCH}/${TARGET} .endif .for _t in ${_t} .if empty(KNOWN_ARCHES:M${_t}) .error Unknown target ${TARGET_ARCH}:${TARGET}. .endif .endfor .if ${TARGET} == ${MACHINE} TARGET_CPUTYPE?=${CPUTYPE} .else TARGET_CPUTYPE?= .endif .if !empty(TARGET_CPUTYPE) _TARGET_CPUTYPE=${TARGET_CPUTYPE} .else _TARGET_CPUTYPE=dummy .endif _CPUTYPE!= MAKEFLAGS= CPUTYPE=${_TARGET_CPUTYPE} ${MAKE} \ -f /dev/null -m ${.CURDIR}/share/mk -V CPUTYPE .if ${_CPUTYPE} != ${_TARGET_CPUTYPE} .error CPUTYPE global should be set with ?=. .endif .if make(buildworld) BUILD_ARCH!= uname -p .if ${MACHINE_ARCH} != ${BUILD_ARCH} .error To cross-build, set TARGET_ARCH. .endif .endif .if ${MACHINE} == ${TARGET} && ${MACHINE_ARCH} == ${TARGET_ARCH} && !defined(CROSS_BUILD_TESTING) OBJTREE= ${MAKEOBJDIRPREFIX} .else OBJTREE= ${MAKEOBJDIRPREFIX}/${TARGET}.${TARGET_ARCH} .endif WORLDTMP= ${OBJTREE}${.CURDIR}/tmp BPATH= ${WORLDTMP}/legacy/usr/sbin:${WORLDTMP}/legacy/usr/bin:${WORLDTMP}/legacy/bin XPATH= ${WORLDTMP}/usr/sbin:${WORLDTMP}/usr/bin STRICTTMPPATH= ${BPATH}:${XPATH} TMPPATH= ${STRICTTMPPATH}:${PATH} # # Avoid running mktemp(1) unless actually needed. # It may not be functional, e.g., due to new ABI # when in the middle of installing over this system. # .if make(distributeworld) || make(installworld) INSTALLTMP!= /usr/bin/mktemp -d -u -t install .endif # # Building a world goes through the following stages # # 1. legacy stage [BMAKE] # This stage is responsible for creating compatibility # shims that are needed by the bootstrap-tools, # build-tools and cross-tools stages. These are generally # APIs that tools from one of those three stages need to # build that aren't present on the host. # 1. bootstrap-tools stage [BMAKE] # This stage is responsible for creating programs that # are needed for backward compatibility reasons. They # are not built as cross-tools. # 2. build-tools stage [TMAKE] # This stage is responsible for creating the object # tree and building any tools that are needed during # the build process. # 3. cross-tools stage [XMAKE] # This stage is responsible for creating any tools that # are needed for building the system. A cross-compiler is one # of them. # 4. world stage [WMAKE] # This stage actually builds the world. # 5. install stage (optional) [IMAKE] # This stage installs a previously built world. # BOOTSTRAPPING?= 0 # Common environment for world related stages CROSSENV= MAKEOBJDIRPREFIX=${OBJTREE} \ MACHINE_ARCH=${TARGET_ARCH} \ MACHINE=${TARGET} \ CPUTYPE=${TARGET_CPUTYPE} .if ${MK_GROFF} != "no" CROSSENV+= GROFF_BIN_PATH=${WORLDTMP}/legacy/usr/bin \ GROFF_FONT_PATH=${WORLDTMP}/legacy/usr/share/groff_font \ GROFF_TMAC_PATH=${WORLDTMP}/legacy/usr/share/tmac .endif .if defined(TARGET_CFLAGS) CROSSENV+= ${TARGET_CFLAGS} .endif # bootstrap-tools stage BMAKEENV= INSTALL="sh ${.CURDIR}/tools/install.sh" \ PATH=${BPATH}:${PATH} \ WORLDTMP=${WORLDTMP} \ VERSION="${VERSION}" \ MAKEFLAGS="-m ${.CURDIR}/tools/build/mk ${.MAKEFLAGS}" BMAKE= MAKEOBJDIRPREFIX=${WORLDTMP} \ ${BMAKEENV} ${MAKE} ${WORLD_FLAGS} -f Makefile.inc1 \ DESTDIR= \ BOOTSTRAPPING=${OSRELDATE} \ SSP_CFLAGS= \ MK_HTML=no NO_LINT=yes MK_MAN=no \ -DNO_PIC MK_PROFILE=no -DNO_SHARED \ -DNO_CPU_CFLAGS MK_WARNS=no MK_CTF=no \ MK_CLANG_EXTRAS=no MK_CLANG_FULL=no \ MK_LLDB=no MK_TESTS=no \ MK_INCLUDES=yes # build-tools stage TMAKE= MAKEOBJDIRPREFIX=${OBJTREE} \ ${BMAKEENV} ${MAKE} ${WORLD_FLAGS} -f Makefile.inc1 \ TARGET=${TARGET} TARGET_ARCH=${TARGET_ARCH} \ DESTDIR= \ BOOTSTRAPPING=${OSRELDATE} \ SSP_CFLAGS= \ -DNO_LINT \ -DNO_CPU_CFLAGS MK_WARNS=no MK_CTF=no \ MK_CLANG_EXTRAS=no MK_CLANG_FULL=no \ MK_LLDB=no MK_TESTS=no # cross-tools stage XMAKE= TOOLS_PREFIX=${WORLDTMP} ${BMAKE} \ TARGET=${TARGET} TARGET_ARCH=${TARGET_ARCH} \ MK_GDB=no MK_TESTS=no # kernel-tools stage KTMAKEENV= INSTALL="sh ${.CURDIR}/tools/install.sh" \ PATH=${BPATH}:${PATH} \ WORLDTMP=${WORLDTMP} \ VERSION="${VERSION}" KTMAKE= TOOLS_PREFIX=${WORLDTMP} MAKEOBJDIRPREFIX=${WORLDTMP} \ ${KTMAKEENV} ${MAKE} ${WORLD_FLAGS} -f Makefile.inc1 \ DESTDIR= \ BOOTSTRAPPING=${OSRELDATE} \ SSP_CFLAGS= \ MK_HTML=no -DNO_LINT MK_MAN=no \ -DNO_PIC MK_PROFILE=no -DNO_SHARED \ -DNO_CPU_CFLAGS MK_WARNS=no MK_CTF=no # world stage WMAKEENV= ${CROSSENV} \ _SHLIBDIRPREFIX=${WORLDTMP} \ _LDSCRIPTROOT= \ VERSION="${VERSION}" \ INSTALL="sh ${.CURDIR}/tools/install.sh" \ PATH=${TMPPATH} # make hierarchy HMAKE= PATH=${TMPPATH} ${MAKE} LOCAL_MTREE=${LOCAL_MTREE:Q} .if defined(NO_ROOT) HMAKE+= PATH=${TMPPATH} METALOG=${METALOG} -DNO_ROOT .endif .if ${MK_CDDL} == "no" WMAKEENV+= MK_CTF=no .endif .if defined(CROSS_TOOLCHAIN) LOCALBASE?= /usr/local .include "${LOCALBASE}/share/toolchains/${CROSS_TOOLCHAIN}.mk" .endif .if defined(CROSS_TOOLCHAIN_PREFIX) CROSS_COMPILER_PREFIX?=${CROSS_TOOLCHAIN_PREFIX} CROSS_BINUTILS_PREFIX?=${CROSS_TOOLCHAIN_PREFIX} .endif # If we do not have a bootstrap binutils (because the in-tree one does not # support the target architecture), provide a default cross-binutils prefix. # This allows aarch64 builds, for example, to automatically use the # aarch64-binutils port or package. .if !empty(BROKEN_OPTIONS:MBINUTILS_BOOTSTRAP) && \ !defined(CROSS_BINUTILS_PREFIX) CROSS_BINUTILS_PREFIX=/usr/local/${TARGET_ARCH}-freebsd/bin/ .if !exists(${CROSS_BINUTILS_PREFIX}) .error In-tree binutils does not support the ${TARGET_ARCH} architecture. Install the ${TARGET_ARCH}-binutils port or package or set CROSS_BINUTILS_PREFIX. .endif .endif XCOMPILERS= CC CXX CPP .for COMPILER in ${XCOMPILERS} .if defined(CROSS_COMPILER_PREFIX) X${COMPILER}?= ${CROSS_COMPILER_PREFIX}${${COMPILER}} .else X${COMPILER}?= ${${COMPILER}} .endif .endfor XBINUTILS= AS AR LD NM OBJCOPY OBJDUMP RANLIB SIZE STRINGS .for BINUTIL in ${XBINUTILS} .if defined(CROSS_BINUTILS_PREFIX) && \ exists(${CROSS_BINUTILS_PREFIX}${${BINUTIL}}) X${BINUTIL}?= ${CROSS_BINUTILS_PREFIX}${${BINUTIL}} .else X${BINUTIL}?= ${${BINUTIL}} .endif .endfor WMAKEENV+= CC="${XCC} ${XCFLAGS}" CXX="${XCXX} ${XCFLAGS} ${XCXXFLAGS}" \ DEPFLAGS="${DEPFLAGS}" \ CPP="${XCPP} ${XCFLAGS}" \ AS="${XAS}" AR="${XAR}" LD="${XLD}" NM=${XNM} \ OBJDUMP=${XOBJDUMP} OBJCOPY="${XOBJCOPY}" \ RANLIB=${XRANLIB} STRINGS=${XSTRINGS} \ SIZE="${XSIZE}" .if ${XCC:M/*} .if defined(CROSS_BINUTILS_PREFIX) # In the case of xdev-build tools, CROSS_BINUTILS_PREFIX won't be a # directory, but the compiler will look in the right place for it's # tools so we don't need to tell it where to look. .if exists(${CROSS_BINUTILS_PREFIX}) BFLAGS+= -B${CROSS_BINUTILS_PREFIX} .endif .else BFLAGS+= -B${WORLDTMP}/usr/bin .endif .if ${TARGET} == "arm" .if ${TARGET_ARCH:M*hf*} != "" TARGET_ABI= gnueabihf .else TARGET_ABI= gnueabi .endif .endif .if defined(X_COMPILER_TYPE) && ${X_COMPILER_TYPE} == gcc XCFLAGS+= -isystem ${WORLDTMP}/usr/include -L${WORLDTMP}/usr/lib XCXXFLAGS+= -I${WORLDTMP}/usr/include/c++/v1 -std=gnu++11 -L${WORLDTMP}/../lib/libc++ DEPFLAGS+= -I${WORLDTMP}/usr/include/c++/v1 .else TARGET_ABI?= unknown TARGET_TRIPLE?= ${TARGET_ARCH:C/amd64/x86_64/}-${TARGET_ABI}-freebsd11.0 XCFLAGS+= -target ${TARGET_TRIPLE} .endif XCFLAGS+= --sysroot=${WORLDTMP} ${BFLAGS} XCXXFLAGS+= --sysroot=${WORLDTMP} ${BFLAGS} .else .if defined(CROSS_BINUTILS_PREFIX) && exists(${CROSS_BINUTILS_PREFIX}) BFLAGS+= -B${CROSS_BINUTILS_PREFIX} XCFLAGS+= ${BFLAGS} XCXXFLAGS+= ${BFLAGS} .endif .endif # ${XCC:M/*} WMAKE= ${WMAKEENV} ${MAKE} ${WORLD_FLAGS} -f Makefile.inc1 DESTDIR=${WORLDTMP} .if ${TARGET_ARCH} == "amd64" || ${TARGET_ARCH} == "powerpc64" # 32 bit world LIB32_OBJTREE= ${OBJTREE}${.CURDIR}/world32 LIB32TMP= ${OBJTREE}${.CURDIR}/lib32 .if ${TARGET_ARCH} == "amd64" .if empty(TARGET_CPUTYPE) LIB32CPUFLAGS= -march=i686 -mmmx -msse -msse2 .else LIB32CPUFLAGS= -march=${TARGET_CPUTYPE} .endif LIB32WMAKEENV= MACHINE=i386 MACHINE_ARCH=i386 \ MACHINE_CPU="i686 mmx sse sse2" LIB32WMAKEFLAGS= \ AS="${XAS} --32" \ LD="${XLD} -m elf_i386_fbsd -Y P,${LIB32TMP}/usr/lib32" \ OBJCOPY="${XOBJCOPY}" .elif ${TARGET_ARCH} == "powerpc64" .if empty(TARGET_CPUTYPE) LIB32CPUFLAGS= -mcpu=powerpc .else LIB32CPUFLAGS= -mcpu=${TARGET_CPUTYPE} .endif LIB32WMAKEENV= MACHINE=powerpc MACHINE_ARCH=powerpc LIB32WMAKEFLAGS= \ LD="${XLD} -m elf32ppc_fbsd" \ OBJCOPY="${XOBJCOPY}" .endif LIB32FLAGS= -m32 ${LIB32CPUFLAGS} -DCOMPAT_32BIT \ -isystem ${LIB32TMP}/usr/include/ \ -L${LIB32TMP}/usr/lib32 \ -B${LIB32TMP}/usr/lib32 .if ${XCC:M/*} LIB32FLAGS+= --sysroot=${WORLDTMP} .endif # Yes, the flags are redundant. LIB32WMAKEENV+= MAKEOBJDIRPREFIX=${LIB32_OBJTREE} \ _SHLIBDIRPREFIX=${LIB32TMP} \ _LDSCRIPTROOT=${LIB32TMP} \ VERSION="${VERSION}" \ INSTALL="sh ${.CURDIR}/tools/install.sh" \ PATH=${TMPPATH} \ LIBDIR=/usr/lib32 \ SHLIBDIR=/usr/lib32 \ LIBPRIVATEDIR=/usr/lib32/private \ DTRACE="${DTRACE} -32" LIB32WMAKEFLAGS+= CC="${XCC} ${LIB32FLAGS}" \ CXX="${XCXX} ${LIB32FLAGS}" \ DESTDIR=${LIB32TMP} \ -DCOMPAT_32BIT \ -DLIBRARIES_ONLY \ -DNO_CPU_CFLAGS \ MK_CTF=no \ -DNO_LINT \ MK_TESTS=no LIB32WMAKE= ${LIB32WMAKEENV} ${MAKE} ${LIB32WMAKEFLAGS} \ MK_MAN=no MK_HTML=no LIB32IMAKE= ${LIB32WMAKE:NINSTALL=*:NDESTDIR=*:N_LDSCRIPTROOT=*} \ MK_TOOLCHAIN=no ${IMAKE_INSTALL} .endif IMAKEENV= ${CROSSENV:N_LDSCRIPTROOT=*} IMAKE= ${IMAKEENV} ${MAKE} -f Makefile.inc1 \ ${IMAKE_INSTALL} ${IMAKE_MTREE} .if empty(.MAKEFLAGS:M-n) IMAKEENV+= PATH=${STRICTTMPPATH}:${INSTALLTMP} \ LD_LIBRARY_PATH=${INSTALLTMP} \ PATH_LOCALE=${INSTALLTMP}/locale IMAKE+= __MAKE_SHELL=${INSTALLTMP}/sh .else IMAKEENV+= PATH=${TMPPATH}:${INSTALLTMP} .endif .if defined(DB_FROM_SRC) INSTALLFLAGS+= -N ${.CURDIR}/etc MTREEFLAGS+= -N ${.CURDIR}/etc .endif _INSTALL_DDIR= ${DESTDIR}/${DISTDIR} INSTALL_DDIR= ${_INSTALL_DDIR:S://:/:g:C:/$::} .if defined(NO_ROOT) METALOG?= ${DESTDIR}/${DISTDIR}/METALOG IMAKE+= -DNO_ROOT METALOG=${METALOG} INSTALLFLAGS+= -U -M ${METALOG} -D ${INSTALL_DDIR} MTREEFLAGS+= -W .endif .if defined(DB_FROM_SRC) || defined(NO_ROOT) IMAKE_INSTALL= INSTALL="install ${INSTALLFLAGS}" IMAKE_MTREE= MTREE_CMD="mtree ${MTREEFLAGS}" .endif # kernel stage KMAKEENV= ${WMAKEENV} KMAKE= ${KMAKEENV} ${MAKE} ${.MAKEFLAGS} ${KERNEL_FLAGS} KERNEL=${INSTKERNNAME} # # buildworld # # Attempt to rebuild the entire system, with reasonable chance of # success, regardless of how old your existing system is. # _worldtmp: .if ${.CURDIR:C/[^,]//g} != "" # The m4 build of sendmail files doesn't like it if ',' is used # anywhere in the path of it's files. @echo @echo "*** Error: path to source tree contains a comma ','" @echo false .endif @echo @echo "--------------------------------------------------------------" @echo ">>> Rebuilding the temporary build tree" @echo "--------------------------------------------------------------" .if !defined(NO_CLEAN) rm -rf ${WORLDTMP} .if defined(LIB32TMP) rm -rf ${LIB32TMP} .endif .else rm -rf ${WORLDTMP}/legacy/usr/include # XXX - These three can depend on any header file. rm -f ${OBJTREE}${.CURDIR}/usr.bin/kdump/ioctl.c rm -f ${OBJTREE}${.CURDIR}/usr.bin/kdump/kdump_subr.c rm -f ${OBJTREE}${.CURDIR}/usr.bin/truss/ioctl.c .endif .for _dir in \ lib usr legacy/bin legacy/usr mkdir -p ${WORLDTMP}/${_dir} .endfor mtree -deU -f ${.CURDIR}/etc/mtree/BSD.usr.dist \ -p ${WORLDTMP}/legacy/usr >/dev/null .if ${MK_GROFF} != "no" mtree -deU -f ${.CURDIR}/etc/mtree/BSD.groff.dist \ -p ${WORLDTMP}/legacy/usr >/dev/null .endif mtree -deU -f ${.CURDIR}/etc/mtree/BSD.usr.dist \ -p ${WORLDTMP}/usr >/dev/null mtree -deU -f ${.CURDIR}/etc/mtree/BSD.include.dist \ -p ${WORLDTMP}/usr/include >/dev/null ln -sf ${.CURDIR}/sys ${WORLDTMP} .if ${MK_DEBUG_FILES} != "no" # We could instead disable debug files for these build stages mtree -deU -f ${.CURDIR}/etc/mtree/BSD.debug.dist \ -p ${WORLDTMP}/legacy/usr/lib >/dev/null mtree -deU -f ${.CURDIR}/etc/mtree/BSD.debug.dist \ -p ${WORLDTMP}/usr/lib >/dev/null .endif .if ${MK_TESTS} != "no" mkdir -p ${WORLDTMP}${TESTSBASE} mtree -deU -f ${.CURDIR}/etc/mtree/BSD.tests.dist \ -p ${WORLDTMP}${TESTSBASE} >/dev/null .endif .for _mtree in ${LOCAL_MTREE} mtree -deU -f ${.CURDIR}/${_mtree} -p ${WORLDTMP} > /dev/null .endfor _legacy: @echo @echo "--------------------------------------------------------------" @echo ">>> stage 1.1: legacy release compatibility shims" @echo "--------------------------------------------------------------" ${_+_}cd ${.CURDIR}; ${BMAKE} legacy _bootstrap-tools: @echo @echo "--------------------------------------------------------------" @echo ">>> stage 1.2: bootstrap tools" @echo "--------------------------------------------------------------" ${_+_}cd ${.CURDIR}; ${BMAKE} bootstrap-tools _cleanobj: .if !defined(NO_CLEAN) @echo @echo "--------------------------------------------------------------" @echo ">>> stage 2.1: cleaning up the object tree" @echo "--------------------------------------------------------------" ${_+_}cd ${.CURDIR}; ${WMAKE} ${CLEANDIR:S/^/par-/} .if defined(LIB32TMP) ${_+_}cd ${.CURDIR}; ${LIB32WMAKE} -f Makefile.inc1 ${CLEANDIR:S/^/par-/} .endif .endif _obj: @echo @echo "--------------------------------------------------------------" @echo ">>> stage 2.2: rebuilding the object tree" @echo "--------------------------------------------------------------" ${_+_}cd ${.CURDIR}; ${WMAKE} par-obj _build-tools: @echo @echo "--------------------------------------------------------------" @echo ">>> stage 2.3: build tools" @echo "--------------------------------------------------------------" ${_+_}cd ${.CURDIR}; ${TMAKE} build-tools _cross-tools: @echo @echo "--------------------------------------------------------------" @echo ">>> stage 3: cross tools" @echo "--------------------------------------------------------------" ${_+_}cd ${.CURDIR}; ${XMAKE} cross-tools ${_+_}cd ${.CURDIR}; ${XMAKE} kernel-tools _includes: @echo @echo "--------------------------------------------------------------" @echo ">>> stage 4.1: building includes" @echo "--------------------------------------------------------------" ${_+_}cd ${.CURDIR}; ${WMAKE} SHARED=symlinks par-includes _libraries: @echo @echo "--------------------------------------------------------------" @echo ">>> stage 4.2: building libraries" @echo "--------------------------------------------------------------" ${_+_}cd ${.CURDIR}; \ ${WMAKE} -DNO_FSCHG MK_HTML=no -DNO_LINT MK_MAN=no \ MK_PROFILE=no MK_TESTS=no MK_TESTS_SUPPORT=${MK_TESTS} libraries _depend: @echo @echo "--------------------------------------------------------------" @echo ">>> stage 4.3: make dependencies" @echo "--------------------------------------------------------------" ${_+_}cd ${.CURDIR}; ${WMAKE} par-depend everything: @echo @echo "--------------------------------------------------------------" @echo ">>> stage 4.4: building everything" @echo "--------------------------------------------------------------" ${_+_}cd ${.CURDIR}; ${WMAKE} par-all .if defined(LIB32TMP) build32: @echo @echo "--------------------------------------------------------------" @echo ">>> stage 5.1: building 32 bit shim libraries" @echo "--------------------------------------------------------------" mkdir -p ${LIB32TMP}/usr/include mtree -deU -f ${.CURDIR}/etc/mtree/BSD.usr.dist \ -p ${LIB32TMP}/usr >/dev/null mtree -deU -f ${.CURDIR}/etc/mtree/BSD.include.dist \ -p ${LIB32TMP}/usr/include >/dev/null .if ${MK_DEBUG_FILES} != "no" mtree -deU -f ${.CURDIR}/etc/mtree/BSD.debug.dist \ -p ${LIB32TMP}/usr/lib >/dev/null .endif mkdir -p ${WORLDTMP} ln -sf ${.CURDIR}/sys ${WORLDTMP} .for _t in obj includes cd ${.CURDIR}/include; ${LIB32WMAKE} DIRPRFX=include/ ${_t} cd ${.CURDIR}/lib; ${LIB32WMAKE} DIRPRFX=lib/ ${_t} .if ${MK_CDDL} != "no" cd ${.CURDIR}/cddl/lib; ${LIB32WMAKE} DIRPRFX=cddl/lib/ ${_t} .endif cd ${.CURDIR}/gnu/lib; ${LIB32WMAKE} DIRPRFX=gnu/lib/ ${_t} .if ${MK_CRYPT} != "no" cd ${.CURDIR}/secure/lib; ${LIB32WMAKE} DIRPRFX=secure/lib/ ${_t} .endif .if ${MK_KERBEROS} != "no" cd ${.CURDIR}/kerberos5/lib; ${LIB32WMAKE} DIRPRFX=kerberos5/lib ${_t} .endif .endfor .for _dir in usr.bin/lex/lib cd ${.CURDIR}/${_dir}; ${LIB32WMAKE} DIRPRFX=${_dir}/ obj .endfor .for _dir in lib/ncurses/ncurses lib/ncurses/ncursesw lib/libmagic cd ${.CURDIR}/${_dir}; \ WORLDTMP=${WORLDTMP} \ MAKEFLAGS="-m ${.CURDIR}/tools/build/mk ${.MAKEFLAGS}" \ MAKEOBJDIRPREFIX=${LIB32_OBJTREE} ${MAKE} SSP_CFLAGS= DESTDIR= \ DIRPRFX=${_dir}/ -DNO_LINT -DNO_CPU_CFLAGS MK_WARNS=no MK_CTF=no \ build-tools .endfor cd ${.CURDIR}; \ ${LIB32WMAKE} -f Makefile.inc1 libraries .for _t in obj depend all cd ${.CURDIR}/libexec/rtld-elf; PROG=ld-elf32.so.1 ${LIB32WMAKE} \ DIRPRFX=libexec/rtld-elf/ ${_t} cd ${.CURDIR}/usr.bin/ldd; PROG=ldd32 ${LIB32WMAKE} \ DIRPRFX=usr.bin/ldd ${_t} .endfor distribute32 install32: cd ${.CURDIR}/lib; ${LIB32IMAKE} ${.TARGET:S/32$//} .if ${MK_CDDL} != "no" cd ${.CURDIR}/cddl/lib; ${LIB32IMAKE} ${.TARGET:S/32$//} .endif cd ${.CURDIR}/gnu/lib; ${LIB32IMAKE} ${.TARGET:S/32$//} .if ${MK_CRYPT} != "no" cd ${.CURDIR}/secure/lib; ${LIB32IMAKE} ${.TARGET:S/32$//} .endif .if ${MK_KERBEROS} != "no" cd ${.CURDIR}/kerberos5/lib; ${LIB32IMAKE} ${.TARGET:S/32$//} .endif cd ${.CURDIR}/libexec/rtld-elf; \ PROG=ld-elf32.so.1 ${LIB32IMAKE} ${.TARGET:S/32$//} cd ${.CURDIR}/usr.bin/ldd; PROG=ldd32 ${LIB32IMAKE} ${.TARGET:S/32$//} .endif WMAKE_TGTS= .if !defined(SUBDIR_OVERRIDE) WMAKE_TGTS+= _worldtmp _legacy _bootstrap-tools .endif WMAKE_TGTS+= _cleanobj _obj _build-tools .if !defined(SUBDIR_OVERRIDE) WMAKE_TGTS+= _cross-tools .endif WMAKE_TGTS+= _includes _libraries _depend everything .if defined(LIB32TMP) && ${MK_LIB32} != "no" WMAKE_TGTS+= build32 .endif buildworld: buildworld_prologue ${WMAKE_TGTS} buildworld_epilogue .ORDER: buildworld_prologue ${WMAKE_TGTS} buildworld_epilogue buildworld_prologue: @echo "--------------------------------------------------------------" @echo ">>> World build started on `LC_ALL=C date`" @echo "--------------------------------------------------------------" buildworld_epilogue: @echo @echo "--------------------------------------------------------------" @echo ">>> World build completed on `LC_ALL=C date`" @echo "--------------------------------------------------------------" # # We need to have this as a target because the indirection between Makefile # and Makefile.inc1 causes the correct PATH to be used, rather than a # modification of the current environment's PATH. In addition, we need # to quote multiword values. # buildenvvars: @echo ${WMAKEENV:Q} .if ${.TARGETS:Mbuildenv} .if ${.MAKEFLAGS:M-j} .error The buildenv target is incompatible with -j .endif .endif buildenv: @echo Entering world for ${TARGET_ARCH}:${TARGET} @cd ${.CURDIR} && env ${WMAKEENV} ${BUILDENV_SHELL} || true TOOLCHAIN_TGTS= ${WMAKE_TGTS:N_depend:Neverything:Nbuild32} toolchain: ${TOOLCHAIN_TGTS} kernel-toolchain: ${TOOLCHAIN_TGTS:N_includes:N_libraries} # # installcheck # # Checks to be sure system is ready for installworld/installkernel. # installcheck: _installcheck_world _installcheck_kernel _installcheck_world: _installcheck_kernel: # # Require DESTDIR to be set if installing for a different architecture or # using the user/group database in the source tree. # .if ${TARGET_ARCH} != ${MACHINE_ARCH} || ${TARGET} != ${MACHINE} || \ defined(DB_FROM_SRC) .if !make(distributeworld) _installcheck_world: __installcheck_DESTDIR _installcheck_kernel: __installcheck_DESTDIR __installcheck_DESTDIR: .if !defined(DESTDIR) || empty(DESTDIR) @echo "ERROR: Please set DESTDIR!"; \ false .endif .endif .endif .if !defined(DB_FROM_SRC) # # Check for missing UIDs/GIDs. # CHECK_UIDS= auditdistd CHECK_GIDS= audit .if ${MK_SENDMAIL} != "no" CHECK_UIDS+= smmsp CHECK_GIDS+= smmsp .endif .if ${MK_PF} != "no" CHECK_UIDS+= proxy CHECK_GIDS+= proxy authpf .endif .if ${MK_UNBOUND} != "no" CHECK_UIDS+= unbound CHECK_GIDS+= unbound .endif _installcheck_world: __installcheck_UGID __installcheck_UGID: .for uid in ${CHECK_UIDS} @if ! `id -u ${uid} >/dev/null 2>&1`; then \ echo "ERROR: Required ${uid} user is missing, see /usr/src/UPDATING."; \ false; \ fi .endfor .for gid in ${CHECK_GIDS} @if ! `find / -prune -group ${gid} >/dev/null 2>&1`; then \ echo "ERROR: Required ${gid} group is missing, see /usr/src/UPDATING."; \ false; \ fi .endfor .endif # # Required install tools to be saved in a scratch dir for safety. # .if ${MK_ZONEINFO} != "no" _zoneinfo= zic tzsetup .endif ITOOLS= [ awk cap_mkdb cat chflags chmod chown \ date echo egrep find grep id install ${_install-info} \ ln lockf make mkdir mtree mv pwd_mkdb \ rm sed services_mkdb sh strip sysctl test true uname wc ${_zoneinfo} \ ${LOCAL_ITOOLS} # Needed for share/man .if ${MK_MAN} != "no" ITOOLS+=makewhatis .endif # # distributeworld # # Distributes everything compiled by a `buildworld'. # # installworld # # Installs everything compiled by a 'buildworld'. # # Non-base distributions produced by the base system EXTRA_DISTRIBUTIONS= doc .if defined(LIB32TMP) && ${MK_LIB32} != "no" EXTRA_DISTRIBUTIONS+= lib32 .endif .if ${MK_TESTS} != "no" EXTRA_DISTRIBUTIONS+= tests .endif DEBUG_DISTRIBUTIONS= .if ${MK_DEBUG_FILES} != "no" DEBUG_DISTRIBUTIONS+= base ${EXTRA_DISTRIBUTIONS:S,doc,,:S,tests,,} .endif MTREE_MAGIC?= mtree 2.0 distributeworld installworld: _installcheck_world mkdir -p ${INSTALLTMP} progs=$$(for prog in ${ITOOLS}; do \ if progpath=`which $$prog`; then \ echo $$progpath; \ else \ echo "Required tool $$prog not found in PATH." >&2; \ exit 1; \ fi; \ done); \ libs=$$(ldd -f "%o %p\n" -f "%o %p\n" $$progs 2>/dev/null | sort -u | \ while read line; do \ set -- $$line; \ if [ "$$2 $$3" != "not found" ]; then \ echo $$2; \ else \ echo "Required library $$1 not found." >&2; \ exit 1; \ fi; \ done); \ cp $$libs $$progs ${INSTALLTMP} cp -R $${PATH_LOCALE:-"/usr/share/locale"} ${INSTALLTMP}/locale .if defined(NO_ROOT) echo "#${MTREE_MAGIC}" > ${METALOG} .endif .if make(distributeworld) .for dist in ${EXTRA_DISTRIBUTIONS} -mkdir ${DESTDIR}/${DISTDIR}/${dist} mtree -deU -f ${.CURDIR}/etc/mtree/BSD.root.dist \ -p ${DESTDIR}/${DISTDIR}/${dist} >/dev/null mtree -deU -f ${.CURDIR}/etc/mtree/BSD.usr.dist \ -p ${DESTDIR}/${DISTDIR}/${dist}/usr >/dev/null mtree -deU -f ${.CURDIR}/etc/mtree/BSD.include.dist \ -p ${DESTDIR}/${DISTDIR}/${dist}/usr/include >/dev/null .if ${MK_DEBUG_FILES} != "no" mtree -deU -f ${.CURDIR}/etc/mtree/BSD.debug.dist \ -p ${DESTDIR}/${DISTDIR}/${dist}/usr/lib >/dev/null .endif .if ${MK_TESTS} != "no" && ${dist} == "tests" -mkdir -p ${DESTDIR}/${DISTDIR}/${dist}${TESTSBASE} mtree -deU -f ${.CURDIR}/etc/mtree/BSD.tests.dist \ -p ${DESTDIR}/${DISTDIR}/${dist}${TESTSBASE} >/dev/null .endif .if defined(NO_ROOT) ${IMAKEENV} mtree -C -f ${.CURDIR}/etc/mtree/BSD.root.dist | \ sed -e 's#^\./#./${dist}/#' >> ${METALOG} ${IMAKEENV} mtree -C -f ${.CURDIR}/etc/mtree/BSD.usr.dist | \ sed -e 's#^\./#./${dist}/usr/#' >> ${METALOG} ${IMAKEENV} mtree -C -f ${.CURDIR}/etc/mtree/BSD.include.dist | \ sed -e 's#^\./#./${dist}/usr/include/#' >> ${METALOG} .endif .endfor -mkdir ${DESTDIR}/${DISTDIR}/base cd ${.CURDIR}/etc; ${CROSSENV} PATH=${TMPPATH} ${MAKE} \ METALOG=${METALOG} ${IMAKE_INSTALL} ${IMAKE_MTREE} \ DISTBASE=/base DESTDIR=${DESTDIR}/${DISTDIR}/base \ LOCAL_MTREE=${LOCAL_MTREE:Q} distrib-dirs .endif ${_+_}cd ${.CURDIR}; ${IMAKE} re${.TARGET:S/world$//}; \ ${IMAKEENV} rm -rf ${INSTALLTMP} .if make(distributeworld) .for dist in ${EXTRA_DISTRIBUTIONS} find ${DESTDIR}/${DISTDIR}/${dist} -mindepth 1 -empty -delete .endfor .if defined(NO_ROOT) .for dist in base ${EXTRA_DISTRIBUTIONS} @# For each file that exists in this dist, print the corresponding @# line from the METALOG. This relies on the fact that @# a line containing only the filename will sort immediatly before @# the relevant mtree line. cd ${DESTDIR}/${DISTDIR}; \ find ./${dist} | sort -u ${METALOG} - | \ awk 'BEGIN { print "#${MTREE_MAGIC}" } !/ type=/ { file = $$1 } / type=/ { if ($$1 == file) { sub(/^\.\/${dist}\//, "./"); print } }' > \ ${DESTDIR}/${DISTDIR}/${dist}.meta .endfor .for dist in ${DEBUG_DISTRIBUTIONS} @# For each file that exists in this dist, print the corresponding @# line from the METALOG. This relies on the fact that @# a line containing only the filename will sort immediatly before @# the relevant mtree line. cd ${DESTDIR}/${DISTDIR}; \ find ./${dist}/usr/lib/debug | sort -u ${METALOG} - | \ awk 'BEGIN { print "#${MTREE_MAGIC}" } !/ type=/ { file = $$1 } / type=/ { if ($$1 == file) { sub(/^\.\/${dist}\//, "./"); print } }' > \ ${DESTDIR}/${DISTDIR}/${dist}.debug.meta .endfor .endif .endif packageworld: .for dist in base ${EXTRA_DISTRIBUTIONS} .if defined(NO_ROOT) ${_+_}cd ${DESTDIR}/${DISTDIR}/${dist}; \ tar cvf - --exclude usr/lib/debug \ @${DESTDIR}/${DISTDIR}/${dist}.meta | \ ${XZ_CMD} > ${DESTDIR}/${DISTDIR}/${dist}.txz .else ${_+_}cd ${DESTDIR}/${DISTDIR}/${dist}; \ tar cvf - --exclude usr/lib/debug . | \ ${XZ_CMD} > ${DESTDIR}/${DISTDIR}/${dist}.txz .endif .endfor .for dist in ${DEBUG_DISTRIBUTIONS} . if defined(NO_ROOT) ${_+_}cd ${DESTDIR}/${DISTDIR}/${dist}; \ tar cvf - @${DESTDIR}/${DISTDIR}/${dist}.debug.meta | \ ${XZ_CMD} > ${DESTDIR}/${DISTDIR}/${dist}-dbg.txz . else ${_+_}cd ${DESTDIR}/${DISTDIR}/${dist}; \ tar cvLf - usr/lib/debug | \ ${XZ_CMD} > ${DESTDIR}/${DISTDIR}/${dist}-dbg.txz . endif .endfor # # reinstall # # If you have a build server, you can NFS mount the source and obj directories # and do a 'make reinstall' on the *client* to install new binaries from the # most recent server build. # reinstall: .MAKE @echo "--------------------------------------------------------------" @echo ">>> Making hierarchy" @echo "--------------------------------------------------------------" ${_+_}cd ${.CURDIR}; ${MAKE} -f Makefile.inc1 \ LOCAL_MTREE=${LOCAL_MTREE:Q} hierarchy @echo @echo "--------------------------------------------------------------" @echo ">>> Installing everything" @echo "--------------------------------------------------------------" ${_+_}cd ${.CURDIR}; ${MAKE} -f Makefile.inc1 install .if defined(LIB32TMP) && ${MK_LIB32} != "no" ${_+_}cd ${.CURDIR}; ${MAKE} -f Makefile.inc1 install32 .endif redistribute: .MAKE @echo "--------------------------------------------------------------" @echo ">>> Distributing everything" @echo "--------------------------------------------------------------" ${_+_}cd ${.CURDIR}; ${MAKE} -f Makefile.inc1 distribute .if defined(LIB32TMP) && ${MK_LIB32} != "no" ${_+_}cd ${.CURDIR}; ${MAKE} -f Makefile.inc1 distribute32 \ DISTRIBUTION=lib32 .endif distrib-dirs distribution: .MAKE cd ${.CURDIR}/etc; ${CROSSENV} PATH=${TMPPATH} ${MAKE} \ ${IMAKE_INSTALL} ${IMAKE_MTREE} METALOG=${METALOG} ${.TARGET} # # buildkernel and installkernel # # Which kernels to build and/or install is specified by setting # KERNCONF. If not defined a GENERIC kernel is built/installed. # Only the existing (depending TARGET) config files are used # for building kernels and only the first of these is designated # as the one being installed. # # Note that we have to use TARGET instead of TARGET_ARCH when # we're in kernel-land. Since only TARGET_ARCH is (expected) to # be set to cross-build, we have to make sure TARGET is set # properly. .if defined(KERNFAST) NO_KERNELCLEAN= t NO_KERNELCONFIG= t NO_KERNELDEPEND= t NO_KERNELOBJ= t # Shortcut for KERNCONF=Blah -DKERNFAST is now KERNFAST=Blah .if !defined(KERNCONF) && ${KERNFAST} != "1" KERNCONF=${KERNFAST} .endif .endif .if ${TARGET_ARCH} == "powerpc64" KERNCONF?= GENERIC64 .else KERNCONF?= GENERIC .endif INSTKERNNAME?= kernel KERNSRCDIR?= ${.CURDIR}/sys KRNLCONFDIR= ${KERNSRCDIR}/${TARGET}/conf KRNLOBJDIR= ${OBJTREE}${KERNSRCDIR} KERNCONFDIR?= ${KRNLCONFDIR} BUILDKERNELS= INSTALLKERNEL= .for _kernel in ${KERNCONF} .if exists(${KERNCONFDIR}/${_kernel}) BUILDKERNELS+= ${_kernel} .if empty(INSTALLKERNEL) INSTALLKERNEL= ${_kernel} .endif .endif .endfor buildkernel ${WMAKE_TGTS} ${.ALLTARGETS:M_*}: .MAKE # # buildkernel # # Builds all kernels defined by BUILDKERNELS. # buildkernel: .if empty(BUILDKERNELS) @echo "ERROR: Missing kernel configuration file(s) (${KERNCONF})."; \ false .endif @echo .for _kernel in ${BUILDKERNELS} @echo "--------------------------------------------------------------" @echo ">>> Kernel build for ${_kernel} started on `LC_ALL=C date`" @echo "--------------------------------------------------------------" @echo "===> ${_kernel}" mkdir -p ${KRNLOBJDIR} .if !defined(NO_KERNELCONFIG) @echo @echo "--------------------------------------------------------------" @echo ">>> stage 1: configuring the kernel" @echo "--------------------------------------------------------------" cd ${KRNLCONFDIR}; \ PATH=${TMPPATH} \ config ${CONFIGARGS} -d ${KRNLOBJDIR}/${_kernel} \ -I '${KERNCONFDIR}' '${KERNCONFDIR}/${_kernel}' .endif .if !defined(NO_CLEAN) && !defined(NO_KERNELCLEAN) @echo @echo "--------------------------------------------------------------" @echo ">>> stage 2.1: cleaning up the object tree" @echo "--------------------------------------------------------------" cd ${KRNLOBJDIR}/${_kernel}; ${KMAKE} ${CLEANDIR} .endif .if !defined(NO_KERNELOBJ) @echo @echo "--------------------------------------------------------------" @echo ">>> stage 2.2: rebuilding the object tree" @echo "--------------------------------------------------------------" cd ${KRNLOBJDIR}/${_kernel}; ${KMAKE} obj .endif @echo @echo "--------------------------------------------------------------" @echo ">>> stage 2.3: build tools" @echo "--------------------------------------------------------------" ${_+_}cd ${.CURDIR}; ${KTMAKE} kernel-tools .if !defined(NO_KERNELDEPEND) @echo @echo "--------------------------------------------------------------" @echo ">>> stage 3.1: making dependencies" @echo "--------------------------------------------------------------" cd ${KRNLOBJDIR}/${_kernel}; ${KMAKE} depend -DNO_MODULES_OBJ .endif @echo @echo "--------------------------------------------------------------" @echo ">>> stage 3.2: building everything" @echo "--------------------------------------------------------------" cd ${KRNLOBJDIR}/${_kernel}; ${KMAKE} all -DNO_MODULES_OBJ @echo "--------------------------------------------------------------" @echo ">>> Kernel build for ${_kernel} completed on `LC_ALL=C date`" @echo "--------------------------------------------------------------" .endfor # # installkernel, etc. # # Install the kernel defined by INSTALLKERNEL # installkernel installkernel.debug \ reinstallkernel reinstallkernel.debug: _installcheck_kernel .if empty(INSTALLKERNEL) @echo "ERROR: No kernel \"${KERNCONF}\" to install."; \ false .endif @echo "--------------------------------------------------------------" @echo ">>> Installing kernel ${INSTALLKERNEL}" @echo "--------------------------------------------------------------" cd ${KRNLOBJDIR}/${INSTALLKERNEL}; \ ${CROSSENV} PATH=${TMPPATH} \ ${MAKE} ${IMAKE_INSTALL} KERNEL=${INSTKERNNAME} ${.TARGET:S/kernel//} distributekernel distributekernel.debug: .if empty(INSTALLKERNEL) @echo "ERROR: No kernel \"${KERNCONF}\" to install."; \ false .endif mkdir -p ${DESTDIR}/${DISTDIR} .if defined(NO_ROOT) echo "#${MTREE_MAGIC}" > ${DESTDIR}/${DISTDIR}/kernel.premeta .endif cd ${KRNLOBJDIR}/${INSTALLKERNEL}; \ ${IMAKEENV} ${IMAKE_INSTALL:S/METALOG/kernel.premeta/} \ ${IMAKE_MTREE} PATH=${TMPPATH} ${MAKE} KERNEL=${INSTKERNNAME} \ DESTDIR=${INSTALL_DDIR}/kernel \ ${.TARGET:S/distributekernel/install/} .if defined(NO_ROOT) sed -e 's|^./kernel|.|' ${DESTDIR}/${DISTDIR}/kernel.premeta > \ ${DESTDIR}/${DISTDIR}/kernel.meta .endif .for _kernel in ${BUILDKERNELS:S/${INSTALLKERNEL}//} .if defined(NO_ROOT) echo "#${MTREE_MAGIC}" > ${DESTDIR}/${DISTDIR}/kernel.${_kernel}.premeta .endif cd ${KRNLOBJDIR}/${_kernel}; \ ${IMAKEENV} ${IMAKE_INSTALL:S/METALOG/kernel.${_kernel}.premeta/} \ ${IMAKE_MTREE} PATH=${TMPPATH} ${MAKE} \ KERNEL=${INSTKERNNAME}.${_kernel} \ DESTDIR=${INSTALL_DDIR}/kernel.${_kernel} \ ${.TARGET:S/distributekernel/install/} .if defined(NO_ROOT) sed -e 's|^./kernel|.|' \ ${DESTDIR}/${DISTDIR}/kernel.${_kernel}.premeta > \ ${DESTDIR}/${DISTDIR}/kernel.${_kernel}.meta .endif .endfor packagekernel: .if defined(NO_ROOT) cd ${DESTDIR}/${DISTDIR}/kernel; \ tar cvf - @${DESTDIR}/${DISTDIR}/kernel.meta | \ ${XZ_CMD} > ${DESTDIR}/${DISTDIR}/kernel.txz .for _kernel in ${BUILDKERNELS:S/${INSTALLKERNEL}//} cd ${DESTDIR}/${DISTDIR}/kernel.${_kernel}; \ tar cvf - @${DESTDIR}/${DISTDIR}/kernel.${_kernel}.meta | \ ${XZ_CMD} > ${DESTDIR}/${DISTDIR}/kernel.${_kernel}.txz .endfor .else cd ${DESTDIR}/${DISTDIR}/kernel; \ tar cvf - . | \ ${XZ_CMD} > ${DESTDIR}/${DISTDIR}/kernel.txz .for _kernel in ${BUILDKERNELS:S/${INSTALLKERNEL}//} cd ${DESTDIR}/${DISTDIR}/kernel.${_kernel}; \ tar cvf - . | \ ${XZ_CMD} > ${DESTDIR}/${DISTDIR}/kernel.${_kernel}.txz .endfor .endif # # doxygen # # Build the API documentation with doxygen # doxygen: @if [ ! -x `/usr/bin/which doxygen` ]; then \ echo "You need doxygen (devel/doxygen) to generate the API documentation of the kernel." | /usr/bin/fmt; \ exit 1; \ fi cd ${.CURDIR}/tools/kerneldoc/subsys && ${MAKE} obj all # # update # # Update the source tree(s), by running svn/svnup to update to the # latest copy. # update: .if (defined(CVS_UPDATE) || defined(SUP_UPDATE)) && !defined(SVN_UPDATE) @echo "--------------------------------------------------------------" @echo "CVS_UPDATE and SUP_UPDATE are no longer supported." @echo "Please see: https://wiki.freebsd.org/CvsIsDeprecated" @echo "--------------------------------------------------------------" @exit 1 .endif .if defined(SVN_UPDATE) @echo "--------------------------------------------------------------" @echo ">>> Updating ${.CURDIR} using Subversion" @echo "--------------------------------------------------------------" @(cd ${.CURDIR} && ${SVN} update ${SVNFLAGS}) .endif # # ------------------------------------------------------------------------ # # From here onwards are utility targets used by the 'make world' and # related targets. If your 'world' breaks, you may like to try to fix # the problem and manually run the following targets to attempt to # complete the build. Beware, this is *not* guaranteed to work, you # need to have a pretty good grip on the current state of the system # to attempt to manually finish it. If in doubt, 'make world' again. # # # legacy: Build compatibility shims for the next three targets. This is a minimal # set of tools and shims necessary to compensate for older systems which don't have # the APIs that the targets built in bootstrap-tools, build-tools or cross-tools. # legacy: .if ${BOOTSTRAPPING} < 800107 && ${BOOTSTRAPPING} != 0 @echo "ERROR: Source upgrades from versions prior to 8.0 not supported."; \ false .endif .for _tool in tools/build ${_+_}@${ECHODIR} "===> ${_tool} (obj,includes,depend,all,install)"; \ cd ${.CURDIR}/${_tool} && \ ${MAKE} DIRPRFX=${_tool}/ obj && \ ${MAKE} DIRPRFX=${_tool}/ DESTDIR=${MAKEOBJDIRPREFIX}/legacy includes && \ ${MAKE} DIRPRFX=${_tool}/ depend && \ ${MAKE} DIRPRFX=${_tool}/ all && \ ${MAKE} DIRPRFX=${_tool}/ DESTDIR=${MAKEOBJDIRPREFIX}/legacy install .endfor # # bootstrap-tools: Build tools needed for compatibility. These are binaries that # are built to build other binaries in the system. However, the focus of these # binaries is usually quite narrow. Bootstrap tools use the host's compiler and # libraries, augmented by -legacy. # _bt= _bootstrap-tools .if ${MK_GAMES} != "no" _strfile= games/fortune/strfile .endif .if ${MK_CXX} != "no" _gperf= gnu/usr.bin/gperf .endif .if ${MK_GROFF} != "no" _groff= gnu/usr.bin/groff \ usr.bin/soelim .endif .if ${MK_VT} != "no" _vtfontcvt= usr.bin/vtfontcvt .endif .if ${BOOTSTRAPPING} < 900002 _sed= usr.bin/sed .endif .if ${BOOTSTRAPPING} < 1000002 -_m4= lib/libohash \ - usr.bin/m4 +_libohash= lib/libohash +_m4= usr.bin/m4 ${_bt}-usr.bin/m4: ${_bt}-lib/libohash .endif .if ${BOOTSTRAPPING} < 1000026 _nmtree= lib/libnetbsd \ usr.sbin/nmtree ${_bt}-usr.sbin/nmtree: ${_bt}-lib/libnetbsd .endif .if ${BOOTSTRAPPING} < 1000027 _cat= bin/cat .endif .if ${BOOTSTRAPPING} < 1000033 _lex= usr.bin/lex .endif # r277259 crunchide: Correct 64-bit section header offset # r281674 crunchide: always include both 32- and 64-bit ELF support .if ${BOOTSTRAPPING} < 1100071 _crunch= usr.sbin/crunch .endif .if ${BOOTSTRAPPING} >= 900040 && ${BOOTSTRAPPING} < 900041 _awk= usr.bin/awk .endif _yacc= lib/liby \ usr.bin/yacc ${_bt}-usr.bin/yacc: ${_bt}-lib/liby .if ${MK_BSNMP} != "no" _gensnmptree= usr.sbin/bsnmpd/gensnmptree .endif # We need to build tblgen when we're building clang either as # the bootstrap compiler, or as the part of the normal build. .if ${MK_CLANG_BOOTSTRAP} != "no" || ${MK_CLANG} != "no" _clang_tblgen= \ lib/clang/libllvmsupport \ lib/clang/libllvmtablegen \ usr.bin/clang/tblgen \ usr.bin/clang/clang-tblgen ${_bt}-usr.bin/clang/clang-tblgen: ${_bt}-lib/clang/libllvmtablegen ${_bt}-lib/clang/libllvmsupport ${_bt}-usr.bin/clang/tblgen: ${_bt}-lib/clang/libllvmtablegen ${_bt}-lib/clang/libllvmsupport .endif # ELF Tool Chain libraries are needed for ELF tools and dtrace tools. # dtrace tools are required for older bootstrap env and cross-build # pre libdwarf .if ${BOOTSTRAPPING} < 1100006 || (${MACHINE} != ${TARGET} || \ ${MACHINE_ARCH} != ${TARGET_ARCH}) _elftoolchain_libs= lib/libelf lib/libdwarf .if ${MK_CDDL} != "no" _dtrace_tools= cddl/usr.bin/sgsmsg cddl/lib/libctf cddl/usr.bin/ctfconvert \ cddl/usr.bin/ctfmerge ${_bt}-cddl/usr.bin/ctfconvert: ${_bt}-lib/libelf ${_bt}-lib/libdwarf ${_bt}-cddl/lib/libctf ${_bt}-cddl/usr.bin/ctfmerge: ${_bt}-lib/libelf ${_bt}-lib/libdwarf ${_bt}-cddl/lib/libctf .endif .endif # Default to building the GPL DTC, but build the BSDL one if users explicitly # request it. _dtc= usr.bin/dtc .if ${MK_GPL_DTC} != "no" _dtc= gnu/usr.bin/dtc .endif .if ${MK_KERBEROS} != "no" _kerberos5_bootstrap_tools= \ kerberos5/tools/make-roken \ kerberos5/lib/libroken \ kerberos5/lib/libvers \ kerberos5/tools/asn1_compile \ kerberos5/tools/slc \ usr.bin/compile_et .ORDER: ${_kerberos5_bootstrap_tools:C/^/${_bt}-/g} .endif +.if ${MK_MANDOCDB} != "no" +_libohash?= lib/libohash +_makewhatis= lib/libsqlite3 \ + usr.bin/mandoc +${_bt}-usr.bin/mandoc: ${_bt}-lib/libohash ${_bt}-lib/libsqlite3 +.else +_makewhatis=usr.bin/makewhatis +.endif + # Rebuild up-to-date libmd for xinstall ${_bt}-usr.bin/xinstall: ${_bt}-lib/libmd bootstrap-tools: .PHONY # Please document (add comment) why something is in 'bootstrap-tools'. # Try to bound the building of the bootstrap-tool to just the # FreeBSD versions that need the tool built at this stage of the build. .for _tool in \ ${_clang_tblgen} \ ${_kerberos5_bootstrap_tools} \ ${_elftoolchain_libs} \ ${_dtrace_tools} \ ${_strfile} \ ${_gperf} \ ${_groff} \ ${_dtc} \ ${_awk} \ ${_cat} \ usr.bin/lorder \ - usr.bin/makewhatis \ + ${_libohash} \ + ${_makewhatis} \ usr.bin/rpcgen \ ${_sed} \ ${_yacc} \ ${_m4} \ ${_lex} \ lib/libmd \ usr.bin/xinstall \ ${_gensnmptree} \ usr.sbin/config \ ${_crunch} \ ${_nmtree} \ ${_vtfontcvt} ${_bt}-${_tool}: .PHONY .MAKE ${_+_}@${ECHODIR} "===> ${_tool} (obj,depend,all,install)"; \ cd ${.CURDIR}/${_tool} && \ ${MAKE} DIRPRFX=${_tool}/ obj && \ ${MAKE} DIRPRFX=${_tool}/ depend && \ ${MAKE} DIRPRFX=${_tool}/ all && \ ${MAKE} DIRPRFX=${_tool}/ DESTDIR=${MAKEOBJDIRPREFIX}/legacy install bootstrap-tools: ${_bt}-${_tool} .endfor # # build-tools: Build special purpose build tools # .if !defined(NO_SHARE) _share= share/syscons/scrnmaps .endif .if ${MK_GCC} != "no" _gcc_tools= gnu/usr.bin/cc/cc_tools .endif .if ${MK_RESCUE} != "no" _rescue= rescue/rescue .endif build-tools: .MAKE .for _tool in \ bin/csh \ bin/sh \ ${_rescue} \ ${LOCAL_TOOL_DIRS} \ lib/ncurses/ncurses \ lib/ncurses/ncursesw \ ${_share} \ usr.bin/awk \ lib/libmagic \ usr.bin/mkesdb_static \ usr.bin/mkcsmapper_static \ usr.bin/vi/catalog ${_+_}@${ECHODIR} "===> ${_tool} (obj,build-tools)"; \ cd ${.CURDIR}/${_tool} && \ ${MAKE} DIRPRFX=${_tool}/ obj && \ ${MAKE} DIRPRFX=${_tool}/ build-tools .endfor .for _tool in \ ${_gcc_tools} ${_+_}@${ECHODIR} "===> ${_tool} (obj,depend,all)"; \ cd ${.CURDIR}/${_tool} && \ ${MAKE} DIRPRFX=${_tool}/ obj && \ ${MAKE} DIRPRFX=${_tool}/ depend && \ ${MAKE} DIRPRFX=${_tool}/ all .endfor # # kernel-tools: Build kernel-building tools # kernel-tools: .MAKE mkdir -p ${MAKEOBJDIRPREFIX}/usr mtree -deU -f ${.CURDIR}/etc/mtree/BSD.usr.dist \ -p ${MAKEOBJDIRPREFIX}/usr >/dev/null # # cross-tools: All the tools needed to build the rest of the system after # we get done with the earlier stages. It is the last set of tools needed # to begin building the target binaries. # .if ${TARGET_ARCH} != ${MACHINE_ARCH} .if ${TARGET_ARCH} == "amd64" || ${TARGET_ARCH} == "i386" _btxld= usr.sbin/btxld .endif .endif # If we're given an XAS, don't build binutils. .if ${XAS:M/*} == "" .if ${MK_BINUTILS_BOOTSTRAP} != "no" _binutils= gnu/usr.bin/binutils .endif .if ${MK_ELFTOOLCHAIN_TOOLS} != "no" _elftctools= lib/libelftc \ usr.bin/elfcopy \ usr.bin/nm \ usr.bin/size \ usr.bin/strings # These are not required by the build, but can be useful for developers who # cross-build on a FreeBSD 10 host: _elftctools+= usr.bin/addr2line .endif .elif ${TARGET_ARCH} != ${MACHINE_ARCH} && ${MK_ELFTOOLCHAIN_TOOLS} != "no" # If cross-building with an external binutils we still need to build strip for # the target (for at least crunchide). _elftctools= lib/libelftc \ usr.bin/elfcopy .endif # If an full path to an external cross compiler is given, don't build # a cross compiler. .if ${XCC:M/*} == "" && ${MK_CROSS_COMPILER} != "no" .if ${MK_CLANG_BOOTSTRAP} != "no" _clang= usr.bin/clang _clang_libs= lib/clang .endif .if ${MK_GCC_BOOTSTRAP} != "no" _cc= gnu/usr.bin/cc .endif .endif .if ${MK_USB} != "no" _usb_tools= sys/boot/usb/tools .endif cross-tools: .MAKE .for _tool in \ ${_clang_libs} \ ${_clang} \ ${_binutils} \ ${_elftctools} \ ${_cc} \ ${_btxld} \ ${_crunchide} \ ${_usb_tools} ${_+_}@${ECHODIR} "===> ${_tool} (obj,depend,all,install)"; \ cd ${.CURDIR}/${_tool} && \ ${MAKE} DIRPRFX=${_tool}/ obj && \ ${MAKE} DIRPRFX=${_tool}/ depend && \ ${MAKE} DIRPRFX=${_tool}/ all && \ ${MAKE} DIRPRFX=${_tool}/ DESTDIR=${MAKEOBJDIRPREFIX} install .endfor NXBENV= MAKEOBJDIRPREFIX=${OBJTREE}/nxb \ INSTALL="sh ${.CURDIR}/tools/install.sh" \ VERSION="${VERSION}" \ PATH=${PATH}:${OBJTREE}/gperf_for_gcc/usr/bin NXBMAKE= ${NXBENV} ${MAKE} \ TBLGEN=${OBJTREE}/nxb-bin/usr/bin/tblgen \ CLANG_TBLGEN=${OBJTREE}/nxb-bin/usr/bin/clang-tblgen \ MACHINE=${TARGET} MACHINE_ARCH=${TARGET_ARCH} \ MK_GDB=no MK_TESTS=no \ SSP_CFLAGS= \ MK_HTML=no NO_LINT=yes MK_MAN=no \ -DNO_PIC MK_PROFILE=no -DNO_SHARED \ -DNO_CPU_CFLAGS MK_WARNS=no MK_CTF=no \ MK_CLANG_EXTRAS=no MK_CLANG_FULL=no \ MK_LLDB=no # native-xtools is the current target for qemu-user cross builds of ports # via poudriere and the imgact_binmisc kernel module. # For non-clang enabled targets that are still using the in tree gcc # we must build a gperf binary for one instance of its Makefiles. On # clang-enabled systems, the gperf binary is obsolete. native-xtools: .MAKE .if ${MK_GCC_BOOTSTRAP} != "no" mkdir -p ${OBJTREE}/gperf_for_gcc/usr/bin ${_+_}@${ECHODIR} "===> ${_gperf} (obj,depend,all,install)"; \ cd ${.CURDIR}/${_gperf} && \ ${NXBMAKE} DIRPRFX=${_gperf}/ obj && \ ${NXBMAKE} DIRPRFX=${_gperf}/ depend && \ ${NXBMAKE} DIRPRFX=${_gperf}/ all && \ ${NXBMAKE} DIRPRFX=${_gperf}/ DESTDIR=${OBJTREE}/gperf_for_gcc install .endif mkdir -p ${OBJTREE}/nxb-bin/bin mkdir -p ${OBJTREE}/nxb-bin/sbin mkdir -p ${OBJTREE}/nxb-bin/usr mtree -deU -f ${.CURDIR}/etc/mtree/BSD.usr.dist \ -p ${OBJTREE}/nxb-bin/usr >/dev/null mtree -deU -f ${.CURDIR}/etc/mtree/BSD.include.dist \ -p ${OBJTREE}/nxb-bin/usr/include >/dev/null .for _tool in \ bin/cat \ bin/chmod \ bin/cp \ bin/csh \ bin/echo \ bin/expr \ bin/hostname \ bin/ln \ bin/ls \ bin/mkdir \ bin/mv \ bin/ps \ bin/realpath \ bin/rm \ bin/rmdir \ bin/sh \ bin/sleep \ ${_clang_tblgen} \ usr.bin/ar \ ${_binutils} \ ${_elftctools} \ ${_cc} \ ${_gcc_tools} \ ${_clang_libs} \ ${_clang} \ sbin/md5 \ sbin/sysctl \ gnu/usr.bin/diff \ usr.bin/awk \ usr.bin/basename \ usr.bin/bmake \ usr.bin/bzip2 \ usr.bin/cmp \ usr.bin/dirname \ usr.bin/env \ usr.bin/fetch \ usr.bin/find \ usr.bin/grep \ usr.bin/gzip \ usr.bin/id \ usr.bin/lex \ usr.bin/lorder \ usr.bin/mktemp \ usr.bin/mt \ usr.bin/patch \ usr.bin/sed \ usr.bin/sort \ usr.bin/tar \ usr.bin/touch \ usr.bin/tr \ usr.bin/true \ usr.bin/uniq \ usr.bin/unzip \ usr.bin/xargs \ usr.bin/xinstall \ usr.bin/xz \ usr.bin/yacc \ usr.sbin/chown ${_+_}@${ECHODIR} "===> ${_tool} (obj,depend,all,install)"; \ cd ${.CURDIR}/${_tool} && \ ${NXBMAKE} DIRPRFX=${_tool}/ obj && \ ${NXBMAKE} DIRPRFX=${_tool}/ depend && \ ${NXBMAKE} DIRPRFX=${_tool}/ all && \ ${NXBMAKE} DIRPRFX=${_tool}/ DESTDIR=${OBJTREE}/nxb-bin install .endfor # # hierarchy - ensure that all the needed directories are present # hierarchy hier: .MAKE cd ${.CURDIR}/etc && ${HMAKE} distrib-dirs # # libraries - build all libraries, and install them under ${DESTDIR}. # # The list of libraries with dependents (${_prebuild_libs}) and their # interdependencies (__L) are built automatically by the # ${.CURDIR}/tools/make_libdeps.sh script. # libraries: .MAKE cd ${.CURDIR} && \ ${MAKE} -f Makefile.inc1 _prereq_libs && \ ${MAKE} -f Makefile.inc1 _startup_libs && \ ${MAKE} -f Makefile.inc1 _prebuild_libs && \ ${MAKE} -f Makefile.inc1 _generic_libs # # static libgcc.a prerequisite for shared libc # _prereq_libs= gnu/lib/libssp/libssp_nonshared gnu/lib/libgcc lib/libcompiler_rt # These dependencies are not automatically generated: # # gnu/lib/csu, gnu/lib/libgcc, lib/csu and lib/libc must be built before # all shared libraries for ELF. # _startup_libs= gnu/lib/csu _startup_libs+= lib/csu _startup_libs+= gnu/lib/libgcc _startup_libs+= lib/libcompiler_rt _startup_libs+= lib/libc _startup_libs+= lib/libc_nonshared .if ${MK_LIBCPLUSPLUS} != "no" _startup_libs+= lib/libcxxrt .endif gnu/lib/libgcc__L: lib/libc__L gnu/lib/libgcc__L: lib/libc_nonshared__L .if ${MK_LIBCPLUSPLUS} != "no" lib/libcxxrt__L: gnu/lib/libgcc__L .endif _prebuild_libs= ${_kerberos5_lib_libasn1} \ ${_kerberos5_lib_libhdb} \ ${_kerberos5_lib_libheimbase} \ ${_kerberos5_lib_libheimntlm} \ ${_libsqlite3} \ ${_kerberos5_lib_libheimipcc} \ ${_kerberos5_lib_libhx509} ${_kerberos5_lib_libkrb5} \ ${_kerberos5_lib_libroken} \ ${_kerberos5_lib_libwind} \ lib/libbz2 ${_libcom_err} lib/libcrypt \ lib/libelf lib/libexpat \ lib/libfigpar \ ${_lib_libgssapi} \ lib/libkiconv lib/libkvm lib/liblzma lib/libmd lib/libnv \ ${_lib_libcapsicum} \ lib/ncurses/ncurses lib/ncurses/ncursesw \ lib/libopie lib/libpam ${_lib_libthr} \ ${_lib_libradius} lib/libsbuf lib/libtacplus \ lib/libgeom \ ${_cddl_lib_libumem} ${_cddl_lib_libnvpair} \ ${_cddl_lib_libuutil} \ ${_cddl_lib_libavl} \ ${_cddl_lib_libzfs_core} \ ${_cddl_lib_libctf} \ lib/libutil lib/libpjdlog ${_lib_libypclnt} lib/libz lib/msun \ ${_secure_lib_libcrypto} ${_lib_libldns} \ ${_secure_lib_libssh} ${_secure_lib_libssl} \ gnu/lib/libdialog .if ${MK_GNUCXX} != "no" _prebuild_libs+= gnu/lib/libstdc++ gnu/lib/libsupc++ gnu/lib/libstdc++__L: lib/msun__L gnu/lib/libsupc++__L: gnu/lib/libstdc++__L .endif .if ${MK_LIBCPLUSPLUS} != "no" _prebuild_libs+= lib/libc++ .endif lib/libgeom__L: lib/libexpat__L .if ${MK_LIBTHR} != "no" _lib_libthr= lib/libthr .endif .if ${MK_RADIUS_SUPPORT} != "no" _lib_libradius= lib/libradius .endif .if ${MK_OFED} != "no" _ofed_lib= contrib/ofed/usr.lib/ .endif .if ${MK_CASPER} != "no" _lib_libcapsicum=lib/libcapsicum .endif lib/libcapsicum__L: lib/libnv__L lib/libpjdlog__L: lib/libutil__L lib/liblzma__L: lib/libthr__L _generic_libs= ${_cddl_lib} gnu/lib ${_kerberos5_lib} lib ${_secure_lib} usr.bin/lex/lib ${_ofed_lib} .for _DIR in ${LOCAL_LIB_DIRS} .if exists(${.CURDIR}/${_DIR}/Makefile) _generic_libs+= ${_DIR} .endif .endfor lib/libopie__L lib/libtacplus__L: lib/libmd__L .if ${MK_CDDL} != "no" _cddl_lib_libumem= cddl/lib/libumem _cddl_lib_libnvpair= cddl/lib/libnvpair _cddl_lib_libavl= cddl/lib/libavl _cddl_lib_libuutil= cddl/lib/libuutil _cddl_lib_libzfs_core= cddl/lib/libzfs_core _cddl_lib_libctf= cddl/lib/libctf _cddl_lib= cddl/lib cddl/lib/libzfs_core__L: cddl/lib/libnvpair__L cddl/lib/libzfs__L: lib/libgeom__L cddl/lib/libctf__L: lib/libz__L .endif # cddl/lib/libdtrace requires lib/libproc and lib/librtld_db; it's only built # on select architectures though (see cddl/lib/Makefile) .if ${MACHINE_ARCH} == "amd64" || ${MACHINE_ARCH} == "i386" || \ ${MACHINE_CPUARCH} == "mips" || ${MACHINE_CPUARCH} == "powerpc" || \ ${MACHINE_CPUARCH} == "arm" _prebuild_libs+= lib/libproc lib/librtld_db .endif .if ${MK_CRYPT} != "no" .if ${MK_OPENSSL} != "no" _secure_lib_libcrypto= secure/lib/libcrypto _secure_lib_libssl= secure/lib/libssl lib/libradius__L secure/lib/libssl__L: secure/lib/libcrypto__L .if ${MK_LDNS} != "no" _lib_libldns= lib/libldns lib/libldns__L: secure/lib/libcrypto__L .endif .if ${MK_OPENSSH} != "no" _secure_lib_libssh= secure/lib/libssh secure/lib/libssh__L: lib/libz__L secure/lib/libcrypto__L lib/libcrypt__L .if ${MK_LDNS} != "no" secure/lib/libssh__L: lib/libldns__L .endif .if ${MK_KERBEROS_SUPPORT} != "no" secure/lib/libssh__L: lib/libgssapi__L kerberos5/lib/libkrb5__L \ kerberos5/lib/libhx509__L kerberos5/lib/libasn1__L lib/libcom_err__L \ lib/libmd__L kerberos5/lib/libroken__L .endif .endif .endif _secure_lib= secure/lib .endif .if ${MK_KERBEROS} != "no" kerberos5/lib/libasn1__L: lib/libcom_err__L kerberos5/lib/libroken__L kerberos5/lib/libhdb__L: kerberos5/lib/libasn1__L lib/libcom_err__L \ kerberos5/lib/libkrb5__L kerberos5/lib/libroken__L \ kerberos5/lib/libwind__L lib/libsqlite3__L kerberos5/lib/libheimntlm__L: secure/lib/libcrypto__L kerberos5/lib/libkrb5__L \ kerberos5/lib/libroken__L lib/libcom_err__L kerberos5/lib/libhx509__L: kerberos5/lib/libasn1__L lib/libcom_err__L \ secure/lib/libcrypto__L kerberos5/lib/libroken__L kerberos5/lib/libwind__L kerberos5/lib/libkrb5__L: kerberos5/lib/libasn1__L lib/libcom_err__L \ lib/libcrypt__L secure/lib/libcrypto__L kerberos5/lib/libhx509__L \ kerberos5/lib/libroken__L kerberos5/lib/libwind__L \ kerberos5/lib/libheimbase__L kerberos5/lib/libheimipcc__L kerberos5/lib/libroken__L: lib/libcrypt__L kerberos5/lib/libwind__L: kerberos5/lib/libroken__L lib/libcom_err__L kerberos5/lib/libheimbase__L: lib/libthr__L kerberos5/lib/libheimipcc__L: kerberos5/lib/libroken__L kerberos5/lib/libheimbase__L lib/libthr__L -lib/libsqlite3__L: lib/libthr__L .endif + +lib/libsqlite3__L: lib/libthr__L .if ${MK_GSSAPI} != "no" _lib_libgssapi= lib/libgssapi .endif .if ${MK_KERBEROS} != "no" _kerberos5_lib= kerberos5/lib _kerberos5_lib_libasn1= kerberos5/lib/libasn1 _kerberos5_lib_libhdb= kerberos5/lib/libhdb _kerberos5_lib_libheimbase= kerberos5/lib/libheimbase _kerberos5_lib_libkrb5= kerberos5/lib/libkrb5 _kerberos5_lib_libhx509= kerberos5/lib/libhx509 _kerberos5_lib_libroken= kerberos5/lib/libroken _kerberos5_lib_libheimntlm= kerberos5/lib/libheimntlm _libsqlite3= lib/libsqlite3 _kerberos5_lib_libheimipcc= kerberos5/lib/libheimipcc _kerberos5_lib_libwind= kerberos5/lib/libwind _libcom_err= lib/libcom_err .endif .if ${MK_NIS} != "no" _lib_libypclnt= lib/libypclnt .endif .if ${MK_OPENSSL} == "no" lib/libradius__L: lib/libmd__L .endif lib/libproc__L: \ ${_cddl_lib_libctf:D${_cddl_lib_libctf}__L} lib/libelf__L lib/librtld_db__L lib/libutil__L .if ${MK_CXX} != "no" .if ${MK_LIBCPLUSPLUS} != "no" lib/libproc__L: lib/libcxxrt__L .else # This implies MK_GNUCXX != "no"; see lib/libproc lib/libproc__L: gnu/lib/libsupc++__L .endif .endif gnu/lib/libdialog__L: lib/msun__L lib/ncurses/ncursesw__L .for _lib in ${_prereq_libs} ${_lib}__PL: .PHONY .MAKE .if exists(${.CURDIR}/${_lib}) ${_+_}@${ECHODIR} "===> ${_lib} (obj,depend,all,install)"; \ cd ${.CURDIR}/${_lib} && \ ${MAKE} MK_TESTS=no DIRPRFX=${_lib}/ obj && \ ${MAKE} MK_TESTS=no DIRPRFX=${_lib}/ depend && \ ${MAKE} MK_TESTS=no MK_PROFILE=no -DNO_PIC \ DIRPRFX=${_lib}/ all && \ ${MAKE} MK_TESTS=no MK_PROFILE=no -DNO_PIC \ DIRPRFX=${_lib}/ install .endif .endfor .for _lib in ${_startup_libs} ${_prebuild_libs:Nlib/libpam} ${_generic_libs} ${_lib}__L: .PHONY .MAKE .if exists(${.CURDIR}/${_lib}) ${_+_}@${ECHODIR} "===> ${_lib} (obj,depend,all,install)"; \ cd ${.CURDIR}/${_lib} && \ ${MAKE} MK_TESTS=no DIRPRFX=${_lib}/ obj && \ ${MAKE} MK_TESTS=no DIRPRFX=${_lib}/ depend && \ ${MAKE} MK_TESTS=no DIRPRFX=${_lib}/ all && \ ${MAKE} MK_TESTS=no DIRPRFX=${_lib}/ install .endif .endfor # libpam is special: we need to build static PAM modules before # static PAM library, and dynamic PAM library before dynamic PAM # modules. lib/libpam__L: .PHONY .MAKE ${_+_}@${ECHODIR} "===> lib/libpam (obj,depend,all,install)"; \ cd ${.CURDIR}/lib/libpam && \ ${MAKE} MK_TESTS=no DIRPRFX=lib/libpam/ obj && \ ${MAKE} MK_TESTS=no DIRPRFX=lib/libpam/ depend && \ ${MAKE} MK_TESTS=no DIRPRFX=lib/libpam/ \ -D_NO_LIBPAM_SO_YET all && \ ${MAKE} MK_TESTS=no DIRPRFX=lib/libpam/ \ -D_NO_LIBPAM_SO_YET install _prereq_libs: ${_prereq_libs:S/$/__PL/} _startup_libs: ${_startup_libs:S/$/__L/} _prebuild_libs: ${_prebuild_libs:S/$/__L/} _generic_libs: ${_generic_libs:S/$/__L/} .for __target in all clean cleandepend cleandir depend includes obj .for entry in ${SUBDIR} ${entry}.${__target}__D: .PHONY .MAKE ${_+_}@set -e; if test -d ${.CURDIR}/${entry}.${MACHINE_ARCH}; then \ ${ECHODIR} "===> ${DIRPRFX}${entry}.${MACHINE_ARCH} (${__target})"; \ edir=${entry}.${MACHINE_ARCH}; \ cd ${.CURDIR}/$${edir}; \ else \ ${ECHODIR} "===> ${DIRPRFX}${entry} (${__target})"; \ edir=${entry}; \ cd ${.CURDIR}/$${edir}; \ fi; \ ${MAKE} ${__target} DIRPRFX=${DIRPRFX}$${edir}/ .endfor par-${__target}: ${SUBDIR:S/$/.${__target}__D/} .endfor .include .if make(check-old) || make(check-old-dirs) || \ make(check-old-files) || make(check-old-libs) || \ make(delete-old) || make(delete-old-dirs) || \ make(delete-old-files) || make(delete-old-libs) # # check for / delete old files section # .include "ObsoleteFiles.inc" OLD_LIBS_MESSAGE="Please be sure no application still uses those libraries, \ else you can not start such an application. Consult UPDATING for more \ information regarding how to cope with the removal/revision bump of a \ specific library." .if !defined(BATCH_DELETE_OLD_FILES) RM_I=-i .else RM_I=-v .endif delete-old-files: @echo ">>> Removing old files (only deletes safe to delete libs)" # Ask for every old file if the user really wants to remove it. # It's annoying, but better safe than sorry. # NB: We cannot pass the list of OLD_FILES as a parameter because the # argument list will get too long. Using .for/.endfor make "loops" will make # the Makefile parser segfault. @exec 3<&0; \ cd ${.CURDIR}; \ ${MAKE} -f ${.CURDIR}/Makefile.inc1 ${.MAKEFLAGS} ${.TARGET} \ -V OLD_FILES -V "OLD_FILES:Musr/share/*.gz:R" | xargs -n1 | \ while read file; do \ if [ -f "${DESTDIR}/$${file}" -o -L "${DESTDIR}/$${file}" ]; then \ chflags noschg "${DESTDIR}/$${file}" 2>/dev/null || true; \ rm ${RM_I} "${DESTDIR}/$${file}" <&3; \ fi; \ done # Remove catpages without corresponding manpages. @exec 3<&0; \ find ${DESTDIR}/usr/share/man/cat* ! -type d | \ sed -ep -e's:${DESTDIR}/usr/share/man/cat:${DESTDIR}/usr/share/man/man:' | \ while read catpage; do \ read manpage; \ if [ ! -e "$${manpage}" ]; then \ rm ${RM_I} $${catpage} <&3; \ fi; \ done @echo ">>> Old files removed" check-old-files: @echo ">>> Checking for old files" @cd ${.CURDIR}; \ ${MAKE} -f ${.CURDIR}/Makefile.inc1 ${.MAKEFLAGS} ${.TARGET} \ -V OLD_FILES -V "OLD_FILES:Musr/share/*.gz:R" | xargs -n1 | \ while read file; do \ if [ -f "${DESTDIR}/$${file}" -o -L "${DESTDIR}/$${file}" ]; then \ echo "${DESTDIR}/$${file}"; \ fi; \ done # Check for catpages without corresponding manpages. @find ${DESTDIR}/usr/share/man/cat* ! -type d | \ sed -ep -e's:${DESTDIR}/usr/share/man/cat:${DESTDIR}/usr/share/man/man:' | \ while read catpage; do \ read manpage; \ if [ ! -e "$${manpage}" ]; then \ echo $${catpage}; \ fi; \ done delete-old-libs: @echo ">>> Removing old libraries" @echo "${OLD_LIBS_MESSAGE}" | fmt @exec 3<&0; \ cd ${.CURDIR}; \ ${MAKE} -f ${.CURDIR}/Makefile.inc1 ${.MAKEFLAGS} ${.TARGET} \ -V OLD_LIBS | xargs -n1 | \ while read file; do \ if [ -f "${DESTDIR}/$${file}" -o -L "${DESTDIR}/$${file}" ]; then \ chflags noschg "${DESTDIR}/$${file}" 2>/dev/null || true; \ rm ${RM_I} "${DESTDIR}/$${file}" <&3; \ fi; \ for ext in debug symbols; do \ if ! [ -e "${DESTDIR}/$${file}" ] && [ -f \ "${DESTDIR}${DEBUGDIR}/$${file}.$${ext}" ]; then \ rm ${RM_I} "${DESTDIR}${DEBUGDIR}/$${file}.$${ext}" \ <&3; \ fi; \ done; \ done @echo ">>> Old libraries removed" check-old-libs: @echo ">>> Checking for old libraries" @cd ${.CURDIR}; \ ${MAKE} -f ${.CURDIR}/Makefile.inc1 ${.MAKEFLAGS} ${.TARGET} \ -V OLD_LIBS | xargs -n1 | \ while read file; do \ if [ -f "${DESTDIR}/$${file}" -o -L "${DESTDIR}/$${file}" ]; then \ echo "${DESTDIR}/$${file}"; \ fi; \ for ext in debug symbols; do \ if [ -f "${DESTDIR}${DEBUGDIR}/$${file}.$${ext}" ]; then \ echo "${DESTDIR}${DEBUGDIR}/$${file}.$${ext}"; \ fi; \ done; \ done delete-old-dirs: @echo ">>> Removing old directories" @cd ${.CURDIR}; \ ${MAKE} -f ${.CURDIR}/Makefile.inc1 ${.MAKEFLAGS} ${.TARGET} \ -V OLD_DIRS | xargs -n1 | sort -r | \ while read dir; do \ if [ -d "${DESTDIR}/$${dir}" ]; then \ rmdir -v "${DESTDIR}/$${dir}" || true; \ elif [ -L "${DESTDIR}/$${dir}" ]; then \ echo "${DESTDIR}/$${dir} is a link, please remove everything manually."; \ fi; \ done @echo ">>> Old directories removed" check-old-dirs: @echo ">>> Checking for old directories" @cd ${.CURDIR}; \ ${MAKE} -f ${.CURDIR}/Makefile.inc1 ${.MAKEFLAGS} ${.TARGET} \ -V OLD_DIRS | xargs -n1 | \ while read dir; do \ if [ -d "${DESTDIR}/$${dir}" ]; then \ echo "${DESTDIR}/$${dir}"; \ elif [ -L "${DESTDIR}/$${dir}" ]; then \ echo "${DESTDIR}/$${dir} is a link, please remove everything manually."; \ fi; \ done delete-old: delete-old-files delete-old-dirs @echo "To remove old libraries run '${MAKE} delete-old-libs'." check-old: check-old-files check-old-libs check-old-dirs @echo "To remove old files and directories run '${MAKE} delete-old'." @echo "To remove old libraries run '${MAKE} delete-old-libs'." .endif # # showconfig - show build configuration. # showconfig: @(${MAKE} -n -f ${.CURDIR}/sys/conf/kern.opts.mk -V dummy -dg1; \ ${MAKE} -n -f ${.CURDIR}/share/mk/src.opts.mk -V dummy -dg1) 2>&1 | grep ^MK_ | sort -u .if !empty(KRNLOBJDIR) && !empty(KERNCONF) DTBOUTPUTPATH= ${KRNLOBJDIR}/${KERNCONF}/ .if !defined(FDT_DTS_FILE) || empty(FDT_DTS_FILE) .if exists(${KERNCONFDIR}/${KERNCONF}) FDT_DTS_FILE!= awk 'BEGIN {FS="="} /^makeoptions[[:space:]]+FDT_DTS_FILE/ {print $$2}' \ '${KERNCONFDIR}/${KERNCONF}' ; echo .endif .endif .endif .if !defined(DTBOUTPUTPATH) || !exists(${DTBOUTPUTPATH}) DTBOUTPUTPATH= ${.CURDIR} .endif # # Build 'standalone' Device Tree Blob # builddtb: @PATH=${TMPPATH} MACHINE=${TARGET} \ ${.CURDIR}/sys/tools/fdt/make_dtb.sh ${.CURDIR}/sys \ "${FDT_DTS_FILE}" ${DTBOUTPUTPATH} ############### .if defined(TARGET) && defined(TARGET_ARCH) .if ${TARGET} == ${MACHINE} && ${TARGET_ARCH} == ${MACHINE_ARCH} XDEV_CPUTYPE?=${CPUTYPE} .else XDEV_CPUTYPE?=${TARGET_CPUTYPE} .endif NOFUN=-DNO_FSCHG MK_HTML=no -DNO_LINT \ MK_MAN=no MK_NLS=no MK_PROFILE=no \ MK_KERBEROS=no MK_RESCUE=no MK_TESTS=no MK_WARNS=no \ TARGET=${TARGET} TARGET_ARCH=${TARGET_ARCH} \ CPUTYPE=${XDEV_CPUTYPE} XDDIR=${TARGET_ARCH}-freebsd XDTP?=/usr/${XDDIR} .if ${XDTP:N/*} .error XDTP variable should be an absolute path .endif CDBENV=MAKEOBJDIRPREFIX=${MAKEOBJDIRPREFIX}/${XDDIR} \ INSTALL="sh ${.CURDIR}/tools/install.sh" CDENV= ${CDBENV} \ _SHLIBDIRPREFIX=${XDDESTDIR} \ TOOLS_PREFIX=${XDTP} CD2CFLAGS=-isystem ${XDDESTDIR}/usr/include -L${XDDESTDIR}/usr/lib \ --sysroot=${XDDESTDIR}/ -B${XDDESTDIR}/usr/libexec \ -B${XDDESTDIR}/usr/bin -B${XDDESTDIR}/usr/lib CD2ENV=${CDENV} CC="${CC} ${CD2CFLAGS}" CXX="${CXX} ${CD2CFLAGS}" \ CPP="${CPP} ${CD2CFLAGS}" \ MACHINE=${TARGET} MACHINE_ARCH=${TARGET_ARCH} CDTMP= ${MAKEOBJDIRPREFIX}/${XDDIR}/${.CURDIR}/tmp CDMAKE=${CDENV} PATH=${CDTMP}/usr/bin:${PATH} ${MAKE} ${NOFUN} CD2MAKE=${CD2ENV} PATH=${CDTMP}/usr/bin:${XDDESTDIR}/usr/bin:${PATH} ${MAKE} ${NOFUN} XDDESTDIR=${DESTDIR}/${XDTP} .if !defined(OSREL) OSREL!= uname -r | sed -e 's/[-(].*//' .endif .ORDER: xdev-build xdev-install xdev-links xdev: xdev-build xdev-install .ORDER: _xb-worldtmp _xb-bootstrap-tools _xb-build-tools _xb-cross-tools xdev-build: _xb-worldtmp _xb-bootstrap-tools _xb-build-tools _xb-cross-tools _xb-worldtmp: mkdir -p ${CDTMP}/usr mtree -deU -f ${.CURDIR}/etc/mtree/BSD.usr.dist \ -p ${CDTMP}/usr >/dev/null _xb-bootstrap-tools: .for _tool in \ ${_clang_tblgen} \ ${_gperf} ${_+_}@${ECHODIR} "===> ${_tool} (obj,depend,all,install)"; \ cd ${.CURDIR}/${_tool} && \ ${CDMAKE} DIRPRFX=${_tool}/ obj && \ ${CDMAKE} DIRPRFX=${_tool}/ depend && \ ${CDMAKE} DIRPRFX=${_tool}/ all && \ ${CDMAKE} DIRPRFX=${_tool}/ DESTDIR=${CDTMP} install .endfor _xb-build-tools: ${_+_}@cd ${.CURDIR}; \ ${CDBENV} ${MAKE} -f Makefile.inc1 ${NOFUN} build-tools _xb-cross-tools: .for _tool in \ ${_binutils} \ ${_elftctools} \ usr.bin/ar \ ${_clang_libs} \ ${_clang} \ ${_cc} ${_+_}@${ECHODIR} "===> xdev ${_tool} (obj,depend,all)"; \ cd ${.CURDIR}/${_tool} && \ ${CDMAKE} DIRPRFX=${_tool}/ obj && \ ${CDMAKE} DIRPRFX=${_tool}/ depend && \ ${CDMAKE} DIRPRFX=${_tool}/ all .endfor _xi-mtree: ${_+_}@${ECHODIR} "mtree populating ${XDDESTDIR}" mkdir -p ${XDDESTDIR} mtree -deU -f ${.CURDIR}/etc/mtree/BSD.root.dist \ -p ${XDDESTDIR} >/dev/null mtree -deU -f ${.CURDIR}/etc/mtree/BSD.usr.dist \ -p ${XDDESTDIR}/usr >/dev/null mtree -deU -f ${.CURDIR}/etc/mtree/BSD.include.dist \ -p ${XDDESTDIR}/usr/include >/dev/null .if ${MK_TESTS} != "no" mkdir -p ${XDDESTDIR}${TESTSBASE} mtree -deU -f ${.CURDIR}/etc/mtree/BSD.tests.dist \ -p ${XDDESTDIR}${TESTSBASE} >/dev/null .endif .ORDER: xdev-build _xi-mtree _xi-cross-tools _xi-includes _xi-libraries xdev-install: xdev-build _xi-mtree _xi-cross-tools _xi-includes _xi-libraries _xi-cross-tools: @echo "_xi-cross-tools" .for _tool in \ ${_binutils} \ ${_elftctools} \ usr.bin/ar \ ${_clang_libs} \ ${_clang} \ ${_cc} ${_+_}@${ECHODIR} "===> xdev ${_tool} (install)"; \ cd ${.CURDIR}/${_tool}; \ ${CDMAKE} DIRPRFX=${_tool}/ install DESTDIR=${XDDESTDIR} .endfor _xi-includes: ${_+_}cd ${.CURDIR}; ${CD2MAKE} -f Makefile.inc1 par-includes \ DESTDIR=${XDDESTDIR} _xi-libraries: ${_+_}cd ${.CURDIR}; ${CD2MAKE} -f Makefile.inc1 libraries \ DESTDIR=${XDDESTDIR} xdev-links: ${_+_}cd ${XDDESTDIR}/usr/bin; \ mkdir -p ../../../../usr/bin; \ for i in *; do \ ln -sf ../../${XDTP}/usr/bin/$$i \ ../../../../usr/bin/${XDDIR}-$$i; \ ln -sf ../../${XDTP}/usr/bin/$$i \ ../../../../usr/bin/${XDDIR}${OSREL}-$$i; \ done .else xdev xdev-build xdev-install xdev-links: @echo "*** Error: Both TARGET and TARGET_ARCH must be defined for \"${.TARGET}\" target" .endif Index: projects/clang-trunk/bin/ps/ps.c =================================================================== --- projects/clang-trunk/bin/ps/ps.c (revision 283870) +++ projects/clang-trunk/bin/ps/ps.c (revision 283871) @@ -1,1443 +1,1438 @@ /*- * Copyright (c) 1990, 1993, 1994 * The Regents of the University of California. All rights reserved. * * Redistribution and use in source and binary forms, with or without * modification, are permitted provided that the following conditions * are met: * 1. Redistributions of source code must retain the above copyright * notice, this list of conditions and the following disclaimer. * 2. Redistributions in binary form must reproduce the above copyright * notice, this list of conditions and the following disclaimer in the * documentation and/or other materials provided with the distribution. * 4. Neither the name of the University nor the names of its contributors * may be used to endorse or promote products derived from this software * without specific prior written permission. * * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF * SUCH DAMAGE. * ------+---------+---------+-------- + --------+---------+---------+---------* * Copyright (c) 2004 - Garance Alistair Drosehn . * All rights reserved. * * Significant modifications made to bring `ps' options somewhat closer * to the standard for `ps' as described in SingleUnixSpec-v3. * ------+---------+---------+-------- + --------+---------+---------+---------* */ #ifndef lint static const char copyright[] = "@(#) Copyright (c) 1990, 1993, 1994\n\ The Regents of the University of California. All rights reserved.\n"; #endif /* not lint */ #if 0 #ifndef lint static char sccsid[] = "@(#)ps.c 8.4 (Berkeley) 4/2/94"; #endif /* not lint */ #endif #include __FBSDID("$FreeBSD$"); #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include "ps.h" #define _PATH_PTS "/dev/pts/" #define W_SEP " \t" /* "Whitespace" list separators */ #define T_SEP "," /* "Terminate-element" list separators */ #ifdef LAZY_PS #define DEF_UREAD 0 #define OPT_LAZY_f "f" #else #define DEF_UREAD 1 /* Always do the more-expensive read. */ #define OPT_LAZY_f /* I.e., the `-f' option is not added. */ #endif /* * isdigit takes an `int', but expects values in the range of unsigned char. * This wrapper ensures that values from a 'char' end up in the correct range. */ #define isdigitch(Anychar) isdigit((u_char)(Anychar)) int cflag; /* -c */ int eval; /* Exit value */ time_t now; /* Current time(3) value */ int rawcpu; /* -C */ int sumrusage; /* -S */ int termwidth; /* Width of the screen (0 == infinity). */ int showthreads; /* will threads be shown? */ struct velisthead varlist = STAILQ_HEAD_INITIALIZER(varlist); static int forceuread = DEF_UREAD; /* Do extra work to get u-area. */ static kvm_t *kd; static int needcomm; /* -o "command" */ static int needenv; /* -e */ static int needuser; /* -o "user" */ static int optfatal; /* Fatal error parsing some list-option. */ static int pid_max; /* kern.max_pid */ static enum sort { DEFAULT, SORTMEM, SORTCPU } sortby = DEFAULT; struct listinfo; typedef int addelem_rtn(struct listinfo *_inf, const char *_elem); struct listinfo { int count; int maxcount; int elemsize; addelem_rtn *addelem; const char *lname; union { gid_t *gids; int *jids; pid_t *pids; dev_t *ttys; uid_t *uids; void *ptr; } l; }; static int addelem_gid(struct listinfo *, const char *); static int addelem_jid(struct listinfo *, const char *); static int addelem_pid(struct listinfo *, const char *); static int addelem_tty(struct listinfo *, const char *); static int addelem_uid(struct listinfo *, const char *); static void add_list(struct listinfo *, const char *); static void descendant_sort(KINFO *, int); static void format_output(KINFO *); static void *expand_list(struct listinfo *); static const char * fmt(char **(*)(kvm_t *, const struct kinfo_proc *, int), KINFO *, char *, char *, int); static void free_list(struct listinfo *); static void init_list(struct listinfo *, addelem_rtn, int, const char *); static char *kludge_oldps_options(const char *, char *, const char *); static int pscomp(const void *, const void *); static void saveuser(KINFO *); static void scanvars(void); static void sizevars(void); static void pidmax_init(void); static void usage(void); static char dfmt[] = "pid,tt,state,time,command"; static char jfmt[] = "user,pid,ppid,pgid,sid,jobc,state,tt,time,command"; static char lfmt[] = "uid,pid,ppid,cpu,pri,nice,vsz,rss,mwchan,state," "tt,time,command"; static char o1[] = "pid"; static char o2[] = "tt,state,time,command"; static char ufmt[] = "user,pid,%cpu,%mem,vsz,rss,tt,state,start,time,command"; static char vfmt[] = "pid,state,time,sl,re,pagein,vsz,rss,lim,tsiz," "%cpu,%mem,command"; static char Zfmt[] = "label"; #define PS_ARGS "AaCcde" OPT_LAZY_f "G:gHhjJ:LlM:mN:O:o:p:rSTt:U:uvwXxZ" int main(int argc, char *argv[]) { struct listinfo gidlist, jidlist, pgrplist, pidlist; struct listinfo ruidlist, sesslist, ttylist, uidlist; struct kinfo_proc *kp; KINFO *kinfo = NULL, *next_KINFO; KINFO_STR *ks; struct varent *vent; struct winsize ws = { .ws_row = 0 }; - const char *nlistf, *memf, *fmtstr, *str; + const char *nlistf, *memf, *str; char *cols; int all, ch, elem, flag, _fmt, i, lineno, linelen, left; int descendancy, nentries, nkept, nselectors; int prtheader, wflag, what, xkeep, xkeep_implied; int fwidthmin, fwidthmax; char errbuf[_POSIX2_LINE_MAX]; char fmtbuf[_POSIX2_LINE_MAX]; (void) setlocale(LC_ALL, ""); time(&now); /* Used by routines in print.c. */ if ((cols = getenv("COLUMNS")) != NULL && *cols != '\0') termwidth = atoi(cols); else if ((ioctl(STDOUT_FILENO, TIOCGWINSZ, (char *)&ws) == -1 && ioctl(STDERR_FILENO, TIOCGWINSZ, (char *)&ws) == -1 && ioctl(STDIN_FILENO, TIOCGWINSZ, (char *)&ws) == -1) || ws.ws_col == 0) termwidth = 79; else termwidth = ws.ws_col - 1; /* * Hide a number of option-processing kludges in a separate routine, * to support some historical BSD behaviors, such as `ps axu'. */ if (argc > 1) argv[1] = kludge_oldps_options(PS_ARGS, argv[1], argv[2]); pidmax_init(); all = descendancy = _fmt = nselectors = optfatal = 0; prtheader = showthreads = wflag = xkeep_implied = 0; xkeep = -1; /* Neither -x nor -X. */ init_list(&gidlist, addelem_gid, sizeof(gid_t), "group"); init_list(&jidlist, addelem_jid, sizeof(int), "jail id"); init_list(&pgrplist, addelem_pid, sizeof(pid_t), "process group"); init_list(&pidlist, addelem_pid, sizeof(pid_t), "process id"); init_list(&ruidlist, addelem_uid, sizeof(uid_t), "ruser"); init_list(&sesslist, addelem_pid, sizeof(pid_t), "session id"); init_list(&ttylist, addelem_tty, sizeof(dev_t), "tty"); init_list(&uidlist, addelem_uid, sizeof(uid_t), "user"); memf = _PATH_DEVNULL; nlistf = NULL; argc = xo_parse_args(argc, argv); if (argc < 0) exit(1); while ((ch = getopt(argc, argv, PS_ARGS)) != -1) switch (ch) { case 'A': /* * Exactly the same as `-ax'. This has been * added for compatibility with SUSv3, but for * now it will not be described in the man page. */ nselectors++; all = xkeep = 1; break; case 'a': nselectors++; all = 1; break; case 'C': rawcpu = 1; break; case 'c': cflag = 1; break; case 'd': descendancy = 1; break; case 'e': /* XXX set ufmt */ needenv = 1; break; #ifdef LAZY_PS case 'f': if (getuid() == 0 || getgid() == 0) forceuread = 1; break; #endif case 'G': add_list(&gidlist, optarg); xkeep_implied = 1; nselectors++; break; case 'g': #if 0 /*- * XXX - This SUSv3 behavior is still under debate * since it conflicts with the (undocumented) * `-g' option. So we skip it for now. */ add_list(&pgrplist, optarg); xkeep_implied = 1; nselectors++; break; #else /* The historical BSD-ish (from SunOS) behavior. */ break; /* no-op */ #endif case 'H': showthreads = KERN_PROC_INC_THREAD; break; case 'h': prtheader = ws.ws_row > 5 ? ws.ws_row : 22; break; case 'J': add_list(&jidlist, optarg); xkeep_implied = 1; nselectors++; break; case 'j': parsefmt(jfmt, 0); _fmt = 1; jfmt[0] = '\0'; break; case 'L': showkey(); exit(0); case 'l': parsefmt(lfmt, 0); _fmt = 1; lfmt[0] = '\0'; break; case 'M': memf = optarg; break; case 'm': sortby = SORTMEM; break; case 'N': nlistf = optarg; break; case 'O': parsefmt(o1, 1); parsefmt(optarg, 1); parsefmt(o2, 1); o1[0] = o2[0] = '\0'; _fmt = 1; break; case 'o': parsefmt(optarg, 1); _fmt = 1; break; case 'p': add_list(&pidlist, optarg); /* * Note: `-p' does not *set* xkeep, but any values * from pidlist are checked before xkeep is. That * way they are always matched, even if the user * specifies `-X'. */ nselectors++; break; #if 0 case 'R': /*- * XXX - This un-standard option is still under * debate. This is what SUSv3 defines as * the `-U' option, and while it would be * nice to have, it could cause even more * confusion to implement it as `-R'. */ add_list(&ruidlist, optarg); xkeep_implied = 1; nselectors++; break; #endif case 'r': sortby = SORTCPU; break; case 'S': sumrusage = 1; break; #if 0 case 's': /*- * XXX - This non-standard option is still under * debate. This *is* supported on Solaris, * Linux, and IRIX, but conflicts with `-s' * on NetBSD and maybe some older BSD's. */ add_list(&sesslist, optarg); xkeep_implied = 1; nselectors++; break; #endif case 'T': if ((optarg = ttyname(STDIN_FILENO)) == NULL) xo_errx(1, "stdin: not a terminal"); /* FALLTHROUGH */ case 't': add_list(&ttylist, optarg); xkeep_implied = 1; nselectors++; break; case 'U': /* This is what SUSv3 defines as the `-u' option. */ add_list(&uidlist, optarg); xkeep_implied = 1; nselectors++; break; case 'u': parsefmt(ufmt, 0); sortby = SORTCPU; _fmt = 1; ufmt[0] = '\0'; break; case 'v': parsefmt(vfmt, 0); sortby = SORTMEM; _fmt = 1; vfmt[0] = '\0'; break; case 'w': if (wflag) termwidth = UNLIMITED; else if (termwidth < 131) termwidth = 131; wflag++; break; case 'X': /* * Note that `-X' and `-x' are not standard "selector" * options. For most selector-options, we check *all* * processes to see if any are matched by the given * value(s). After we have a set of all the matched * processes, then `-X' and `-x' govern whether we * modify that *matched* set for processes which do * not have a controlling terminal. `-X' causes * those processes to be deleted from the matched * set, while `-x' causes them to be kept. */ xkeep = 0; break; case 'x': xkeep = 1; break; case 'Z': parsefmt(Zfmt, 0); Zfmt[0] = '\0'; break; case '?': default: usage(); } argc -= optind; argv += optind; /* * If there arguments after processing all the options, attempt * to treat them as a list of process ids. */ while (*argv) { if (!isdigitch(**argv)) break; add_list(&pidlist, *argv); argv++; } if (*argv) { xo_warnx("illegal argument: %s\n", *argv); usage(); } if (optfatal) exit(1); /* Error messages already printed. */ if (xkeep < 0) /* Neither -X nor -x was specified. */ xkeep = xkeep_implied; kd = kvm_openfiles(nlistf, memf, NULL, O_RDONLY, errbuf); if (kd == 0) xo_errx(1, "%s", errbuf); if (!_fmt) parsefmt(dfmt, 0); if (nselectors == 0) { uidlist.l.ptr = malloc(sizeof(uid_t)); if (uidlist.l.ptr == NULL) xo_errx(1, "malloc failed"); nselectors = 1; uidlist.count = uidlist.maxcount = 1; *uidlist.l.uids = getuid(); } /* * scan requested variables, noting what structures are needed, * and adjusting header widths as appropriate. */ scanvars(); /* * Get process list. If the user requested just one selector- * option, then kvm_getprocs can be asked to return just those * processes. Otherwise, have it return all processes, and * then this routine will search that full list and select the * processes which match any of the user's selector-options. */ what = showthreads != 0 ? KERN_PROC_ALL : KERN_PROC_PROC; flag = 0; if (nselectors == 1) { if (gidlist.count == 1) { what = KERN_PROC_RGID | showthreads; flag = *gidlist.l.gids; nselectors = 0; } else if (pgrplist.count == 1) { what = KERN_PROC_PGRP | showthreads; flag = *pgrplist.l.pids; nselectors = 0; } else if (pidlist.count == 1) { what = KERN_PROC_PID | showthreads; flag = *pidlist.l.pids; nselectors = 0; } else if (ruidlist.count == 1) { what = KERN_PROC_RUID | showthreads; flag = *ruidlist.l.uids; nselectors = 0; } else if (sesslist.count == 1) { what = KERN_PROC_SESSION | showthreads; flag = *sesslist.l.pids; nselectors = 0; } else if (ttylist.count == 1) { what = KERN_PROC_TTY | showthreads; flag = *ttylist.l.ttys; nselectors = 0; } else if (uidlist.count == 1) { what = KERN_PROC_UID | showthreads; flag = *uidlist.l.uids; nselectors = 0; } else if (all) { /* No need for this routine to select processes. */ nselectors = 0; } } /* * select procs */ nentries = -1; kp = kvm_getprocs(kd, what, flag, &nentries); if ((kp == NULL && nentries > 0) || (kp != NULL && nentries < 0)) xo_errx(1, "%s", kvm_geterr(kd)); nkept = 0; if (nentries > 0) { if ((kinfo = malloc(nentries * sizeof(*kinfo))) == NULL) xo_errx(1, "malloc failed"); for (i = nentries; --i >= 0; ++kp) { /* * If the user specified multiple selection-criteria, * then keep any process matched by the inclusive OR * of all the selection-criteria given. */ if (pidlist.count > 0) { for (elem = 0; elem < pidlist.count; elem++) if (kp->ki_pid == pidlist.l.pids[elem]) goto keepit; } /* * Note that we had to process pidlist before * filtering out processes which do not have * a controlling terminal. */ if (xkeep == 0) { if ((kp->ki_tdev == NODEV || (kp->ki_flag & P_CONTROLT) == 0)) continue; } if (nselectors == 0) goto keepit; if (gidlist.count > 0) { for (elem = 0; elem < gidlist.count; elem++) if (kp->ki_rgid == gidlist.l.gids[elem]) goto keepit; } if (jidlist.count > 0) { for (elem = 0; elem < jidlist.count; elem++) if (kp->ki_jid == jidlist.l.jids[elem]) goto keepit; } if (pgrplist.count > 0) { for (elem = 0; elem < pgrplist.count; elem++) if (kp->ki_pgid == pgrplist.l.pids[elem]) goto keepit; } if (ruidlist.count > 0) { for (elem = 0; elem < ruidlist.count; elem++) if (kp->ki_ruid == ruidlist.l.uids[elem]) goto keepit; } if (sesslist.count > 0) { for (elem = 0; elem < sesslist.count; elem++) if (kp->ki_sid == sesslist.l.pids[elem]) goto keepit; } if (ttylist.count > 0) { for (elem = 0; elem < ttylist.count; elem++) if (kp->ki_tdev == ttylist.l.ttys[elem]) goto keepit; } if (uidlist.count > 0) { for (elem = 0; elem < uidlist.count; elem++) if (kp->ki_uid == uidlist.l.uids[elem]) goto keepit; } /* * This process did not match any of the user's * selector-options, so skip the process. */ continue; keepit: next_KINFO = &kinfo[nkept]; next_KINFO->ki_p = kp; next_KINFO->ki_d.level = 0; next_KINFO->ki_d.prefix = NULL; next_KINFO->ki_pcpu = getpcpu(next_KINFO); if (sortby == SORTMEM) next_KINFO->ki_memsize = kp->ki_tsize + kp->ki_dsize + kp->ki_ssize; if (needuser) saveuser(next_KINFO); nkept++; } } sizevars(); if (nkept == 0) { printheader(); exit(1); } /* * sort proc list */ qsort(kinfo, nkept, sizeof(KINFO), pscomp); /* * We want things in descendant order */ if (descendancy) descendant_sort(kinfo, nkept); /* * Prepare formatted output. */ for (i = 0; i < nkept; i++) format_output(&kinfo[i]); /* * Print header. */ xo_open_container("process-information"); printheader(); if (xo_get_style(NULL) != XO_STYLE_TEXT) termwidth = UNLIMITED; /* * Output formatted lines. */ xo_open_list("process"); for (i = lineno = 0; i < nkept; i++) { linelen = 0; xo_open_instance("process"); STAILQ_FOREACH(vent, &varlist, next_ve) { - if (vent->var->flag & LJUST) - fmtstr = "%-*s"; - else - fmtstr = "%*s"; - ks = STAILQ_FIRST(&kinfo[i].ki_ks); STAILQ_REMOVE_HEAD(&kinfo[i].ki_ks, ks_next); /* Truncate rightmost column if necessary. */ fwidthmax = _POSIX2_LINE_MAX; if (STAILQ_NEXT(vent, next_ve) == NULL && termwidth != UNLIMITED && ks->ks_str != NULL) { left = termwidth - linelen; if (left > 0 && left < (int)strlen(ks->ks_str)) fwidthmax = left; } str = ks->ks_str; if (str == NULL) str = "-"; /* No padding for the last column, if it's LJUST. */ fwidthmin = (xo_get_style(NULL) != XO_STYLE_TEXT || (STAILQ_NEXT(vent, next_ve) == NULL && (vent->var->flag & LJUST))) ? 0 : vent->var->width; snprintf(fmtbuf, sizeof(fmtbuf), "{:%s/%%%s%d..%ds}", vent->var->field ?: vent->var->name, (vent->var->flag & LJUST) ? "-" : "", fwidthmin, fwidthmax); xo_emit(fmtbuf, str); linelen += fwidthmin; if (ks->ks_str != NULL) { free(ks->ks_str); ks->ks_str = NULL; } free(ks); ks = NULL; if (STAILQ_NEXT(vent, next_ve) != NULL) { xo_emit("{P: }"); linelen++; } } xo_emit("\n"); xo_close_instance("process"); if (prtheader && lineno++ == prtheader - 4) { xo_emit("\n"); printheader(); lineno = 0; } } xo_close_list("process"); xo_close_container("process-information"); xo_finish(); free_list(&gidlist); free_list(&jidlist); free_list(&pidlist); free_list(&pgrplist); free_list(&ruidlist); free_list(&sesslist); free_list(&ttylist); free_list(&uidlist); for (i = 0; i < nkept; i++) free(kinfo[i].ki_d.prefix); free(kinfo); exit(eval); } static int addelem_gid(struct listinfo *inf, const char *elem) { struct group *grp; const char *nameorID; char *endp; u_long bigtemp; if (*elem == '\0' || strlen(elem) >= MAXLOGNAME) { if (*elem == '\0') xo_warnx("Invalid (zero-length) %s name", inf->lname); else xo_warnx("%s name too long: %s", inf->lname, elem); optfatal = 1; return (0); /* Do not add this value. */ } /* * SUSv3 states that `ps -G grouplist' should match "real-group * ID numbers", and does not mention group-names. I do want to * also support group-names, so this tries for a group-id first, * and only tries for a name if that doesn't work. This is the * opposite order of what is done in addelem_uid(), but in * practice the order would only matter for group-names which * are all-numeric. */ grp = NULL; nameorID = "named"; errno = 0; bigtemp = strtoul(elem, &endp, 10); if (errno == 0 && *endp == '\0' && bigtemp <= GID_MAX) { nameorID = "name or ID matches"; grp = getgrgid((gid_t)bigtemp); } if (grp == NULL) grp = getgrnam(elem); if (grp == NULL) { xo_warnx("No %s %s '%s'", inf->lname, nameorID, elem); optfatal = 1; return (0); } if (inf->count >= inf->maxcount) expand_list(inf); inf->l.gids[(inf->count)++] = grp->gr_gid; return (1); } static int addelem_jid(struct listinfo *inf, const char *elem) { int tempid; if (*elem == '\0') { warnx("Invalid (zero-length) jail id"); optfatal = 1; return (0); /* Do not add this value. */ } tempid = jail_getid(elem); if (tempid < 0) { warnx("Invalid %s: %s", inf->lname, elem); optfatal = 1; return (0); } if (inf->count >= inf->maxcount) expand_list(inf); inf->l.jids[(inf->count)++] = tempid; return (1); } static int addelem_pid(struct listinfo *inf, const char *elem) { char *endp; long tempid; if (*elem == '\0') { xo_warnx("Invalid (zero-length) process id"); optfatal = 1; return (0); /* Do not add this value. */ } errno = 0; tempid = strtol(elem, &endp, 10); if (*endp != '\0' || tempid < 0 || elem == endp) { xo_warnx("Invalid %s: %s", inf->lname, elem); errno = ERANGE; } else if (errno != 0 || tempid > pid_max) { xo_warnx("%s too large: %s", inf->lname, elem); errno = ERANGE; } if (errno == ERANGE) { optfatal = 1; return (0); } if (inf->count >= inf->maxcount) expand_list(inf); inf->l.pids[(inf->count)++] = tempid; return (1); } /*- * The user can specify a device via one of three formats: * 1) fully qualified, e.g.: /dev/ttyp0 /dev/console /dev/pts/0 * 2) missing "/dev", e.g.: ttyp0 console pts/0 * 3) two-letters, e.g.: p0 co 0 * (matching letters that would be seen in the "TT" column) */ static int addelem_tty(struct listinfo *inf, const char *elem) { const char *ttypath; struct stat sb; char pathbuf[PATH_MAX], pathbuf2[PATH_MAX], pathbuf3[PATH_MAX]; ttypath = NULL; pathbuf2[0] = '\0'; pathbuf3[0] = '\0'; switch (*elem) { case '/': ttypath = elem; break; case 'c': if (strcmp(elem, "co") == 0) { ttypath = _PATH_CONSOLE; break; } /* FALLTHROUGH */ default: strlcpy(pathbuf, _PATH_DEV, sizeof(pathbuf)); strlcat(pathbuf, elem, sizeof(pathbuf)); ttypath = pathbuf; if (strncmp(pathbuf, _PATH_TTY, strlen(_PATH_TTY)) == 0) break; if (strncmp(pathbuf, _PATH_PTS, strlen(_PATH_PTS)) == 0) break; if (strcmp(pathbuf, _PATH_CONSOLE) == 0) break; /* Check to see if /dev/tty${elem} exists */ strlcpy(pathbuf2, _PATH_TTY, sizeof(pathbuf2)); strlcat(pathbuf2, elem, sizeof(pathbuf2)); if (stat(pathbuf2, &sb) == 0 && S_ISCHR(sb.st_mode)) { /* No need to repeat stat() && S_ISCHR() checks */ ttypath = NULL; break; } /* Check to see if /dev/pts/${elem} exists */ strlcpy(pathbuf3, _PATH_PTS, sizeof(pathbuf3)); strlcat(pathbuf3, elem, sizeof(pathbuf3)); if (stat(pathbuf3, &sb) == 0 && S_ISCHR(sb.st_mode)) { /* No need to repeat stat() && S_ISCHR() checks */ ttypath = NULL; break; } break; } if (ttypath) { if (stat(ttypath, &sb) == -1) { if (pathbuf3[0] != '\0') xo_warn("%s, %s, and %s", pathbuf3, pathbuf2, ttypath); else xo_warn("%s", ttypath); optfatal = 1; return (0); } if (!S_ISCHR(sb.st_mode)) { if (pathbuf3[0] != '\0') xo_warnx("%s, %s, and %s: Not a terminal", pathbuf3, pathbuf2, ttypath); else xo_warnx("%s: Not a terminal", ttypath); optfatal = 1; return (0); } } if (inf->count >= inf->maxcount) expand_list(inf); inf->l.ttys[(inf->count)++] = sb.st_rdev; return (1); } static int addelem_uid(struct listinfo *inf, const char *elem) { struct passwd *pwd; char *endp; u_long bigtemp; if (*elem == '\0' || strlen(elem) >= MAXLOGNAME) { if (*elem == '\0') xo_warnx("Invalid (zero-length) %s name", inf->lname); else xo_warnx("%s name too long: %s", inf->lname, elem); optfatal = 1; return (0); /* Do not add this value. */ } pwd = getpwnam(elem); if (pwd == NULL) { errno = 0; bigtemp = strtoul(elem, &endp, 10); if (errno != 0 || *endp != '\0' || bigtemp > UID_MAX) xo_warnx("No %s named '%s'", inf->lname, elem); else { /* The string is all digits, so it might be a userID. */ pwd = getpwuid((uid_t)bigtemp); if (pwd == NULL) xo_warnx("No %s name or ID matches '%s'", inf->lname, elem); } } if (pwd == NULL) { /* * These used to be treated as minor warnings (and the * option was simply ignored), but now they are fatal * errors (and the command will be aborted). */ optfatal = 1; return (0); } if (inf->count >= inf->maxcount) expand_list(inf); inf->l.uids[(inf->count)++] = pwd->pw_uid; return (1); } static void add_list(struct listinfo *inf, const char *argp) { const char *savep; char *cp, *endp; int toolong; char elemcopy[PATH_MAX]; if (*argp == '\0') inf->addelem(inf, argp); while (*argp != '\0') { while (*argp != '\0' && strchr(W_SEP, *argp) != NULL) argp++; savep = argp; toolong = 0; cp = elemcopy; if (strchr(T_SEP, *argp) == NULL) { endp = elemcopy + sizeof(elemcopy) - 1; while (*argp != '\0' && cp <= endp && strchr(W_SEP T_SEP, *argp) == NULL) *cp++ = *argp++; if (cp > endp) toolong = 1; } if (!toolong) { *cp = '\0'; /* * Add this single element to the given list. */ inf->addelem(inf, elemcopy); } else { /* * The string is too long to copy. Find the end * of the string to print out the warning message. */ while (*argp != '\0' && strchr(W_SEP T_SEP, *argp) == NULL) argp++; xo_warnx("Value too long: %.*s", (int)(argp - savep), savep); optfatal = 1; } /* * Skip over any number of trailing whitespace characters, * but only one (at most) trailing element-terminating * character. */ while (*argp != '\0' && strchr(W_SEP, *argp) != NULL) argp++; if (*argp != '\0' && strchr(T_SEP, *argp) != NULL) { argp++; /* Catch case where string ended with a comma. */ if (*argp == '\0') inf->addelem(inf, argp); } } } static void descendant_sort(KINFO *ki, int items) { int dst, lvl, maxlvl, n, ndst, nsrc, siblings, src; unsigned char *path; KINFO kn; /* * First, sort the entries by descendancy, tracking the descendancy * depth in the ki_d.level field. */ src = 0; maxlvl = 0; while (src < items) { if (ki[src].ki_d.level) { src++; continue; } for (nsrc = 1; src + nsrc < items; nsrc++) if (!ki[src + nsrc].ki_d.level) break; for (dst = 0; dst < items; dst++) { if (ki[dst].ki_p->ki_pid == ki[src].ki_p->ki_pid) continue; if (ki[dst].ki_p->ki_pid == ki[src].ki_p->ki_ppid) break; } if (dst == items) { src += nsrc; continue; } for (ndst = 1; dst + ndst < items; ndst++) if (ki[dst + ndst].ki_d.level <= ki[dst].ki_d.level) break; for (n = src; n < src + nsrc; n++) { ki[n].ki_d.level += ki[dst].ki_d.level + 1; if (maxlvl < ki[n].ki_d.level) maxlvl = ki[n].ki_d.level; } while (nsrc) { if (src < dst) { kn = ki[src]; memmove(ki + src, ki + src + 1, (dst - src + ndst - 1) * sizeof *ki); ki[dst + ndst - 1] = kn; nsrc--; dst--; ndst++; } else if (src != dst + ndst) { kn = ki[src]; memmove(ki + dst + ndst + 1, ki + dst + ndst, (src - dst - ndst) * sizeof *ki); ki[dst + ndst] = kn; ndst++; nsrc--; src++; } else { ndst += nsrc; src += nsrc; nsrc = 0; } } } /* * Now populate ki_d.prefix (instead of ki_d.level) with the command * prefix used to show descendancies. */ path = malloc((maxlvl + 7) / 8); memset(path, '\0', (maxlvl + 7) / 8); for (src = 0; src < items; src++) { if ((lvl = ki[src].ki_d.level) == 0) { ki[src].ki_d.prefix = NULL; continue; } if ((ki[src].ki_d.prefix = malloc(lvl * 2 + 1)) == NULL) xo_errx(1, "malloc failed"); for (n = 0; n < lvl - 2; n++) { ki[src].ki_d.prefix[n * 2] = path[n / 8] & 1 << (n % 8) ? '|' : ' '; ki[src].ki_d.prefix[n * 2 + 1] = ' '; } if (n == lvl - 2) { /* Have I any more siblings? */ for (siblings = 0, dst = src + 1; dst < items; dst++) { if (ki[dst].ki_d.level > lvl) continue; if (ki[dst].ki_d.level == lvl) siblings = 1; break; } if (siblings) path[n / 8] |= 1 << (n % 8); else path[n / 8] &= ~(1 << (n % 8)); ki[src].ki_d.prefix[n * 2] = siblings ? '|' : '`'; ki[src].ki_d.prefix[n * 2 + 1] = '-'; n++; } strcpy(ki[src].ki_d.prefix + n * 2, "- "); } free(path); } static void * expand_list(struct listinfo *inf) { void *newlist; int newmax; newmax = (inf->maxcount + 1) << 1; newlist = realloc(inf->l.ptr, newmax * inf->elemsize); if (newlist == NULL) { free(inf->l.ptr); xo_errx(1, "realloc to %d %ss failed", newmax, inf->lname); } inf->maxcount = newmax; inf->l.ptr = newlist; return (newlist); } static void free_list(struct listinfo *inf) { inf->count = inf->elemsize = inf->maxcount = 0; if (inf->l.ptr != NULL) free(inf->l.ptr); inf->addelem = NULL; inf->lname = NULL; inf->l.ptr = NULL; } static void init_list(struct listinfo *inf, addelem_rtn artn, int elemsize, const char *lname) { inf->count = inf->maxcount = 0; inf->elemsize = elemsize; inf->addelem = artn; inf->lname = lname; inf->l.ptr = NULL; } VARENT * find_varentry(VAR *v) { struct varent *vent; STAILQ_FOREACH(vent, &varlist, next_ve) { if (strcmp(vent->var->name, v->name) == 0) return vent; } return NULL; } static void scanvars(void) { struct varent *vent; VAR *v; STAILQ_FOREACH(vent, &varlist, next_ve) { v = vent->var; if (v->flag & USER) needuser = 1; if (v->flag & COMM) needcomm = 1; } } static void format_output(KINFO *ki) { struct varent *vent; VAR *v; KINFO_STR *ks; char *str; int len; STAILQ_INIT(&ki->ki_ks); STAILQ_FOREACH(vent, &varlist, next_ve) { v = vent->var; str = (v->oproc)(ki, vent); ks = malloc(sizeof(*ks)); if (ks == NULL) xo_errx(1, "malloc failed"); ks->ks_str = str; STAILQ_INSERT_TAIL(&ki->ki_ks, ks, ks_next); if (str != NULL) { len = strlen(str); } else len = 1; /* "-" */ if (v->width < len) v->width = len; } } static void sizevars(void) { struct varent *vent; VAR *v; int i; STAILQ_FOREACH(vent, &varlist, next_ve) { v = vent->var; i = strlen(vent->header); if (v->width < i) v->width = i; } } static const char * fmt(char **(*fn)(kvm_t *, const struct kinfo_proc *, int), KINFO *ki, char *comm, char *thread, int maxlen) { const char *s; s = fmt_argv((*fn)(kd, ki->ki_p, termwidth), comm, showthreads && ki->ki_p->ki_numthreads > 1 ? thread : NULL, maxlen); return (s); } #define UREADOK(ki) (forceuread || (ki->ki_p->ki_flag & P_INMEM)) static void saveuser(KINFO *ki) { if (ki->ki_p->ki_flag & P_INMEM) { /* * The u-area might be swapped out, and we can't get * at it because we have a crashdump and no swap. * If it's here fill in these fields, otherwise, just * leave them 0. */ ki->ki_valid = 1; } else ki->ki_valid = 0; /* * save arguments if needed */ if (needcomm) { if (ki->ki_p->ki_stat == SZOMB) ki->ki_args = strdup(""); else if (UREADOK(ki) || (ki->ki_p->ki_args != NULL)) ki->ki_args = strdup(fmt(kvm_getargv, ki, ki->ki_p->ki_comm, ki->ki_p->ki_tdname, MAXCOMLEN)); else asprintf(&ki->ki_args, "(%s)", ki->ki_p->ki_comm); if (ki->ki_args == NULL) xo_errx(1, "malloc failed"); } else { ki->ki_args = NULL; } if (needenv) { if (UREADOK(ki)) ki->ki_env = strdup(fmt(kvm_getenvv, ki, (char *)NULL, (char *)NULL, 0)); else ki->ki_env = strdup("()"); if (ki->ki_env == NULL) xo_errx(1, "malloc failed"); } else { ki->ki_env = NULL; } } /* A macro used to improve the readability of pscomp(). */ #define DIFF_RETURN(a, b, field) do { \ if ((a)->field != (b)->field) \ return (((a)->field < (b)->field) ? -1 : 1); \ } while (0) static int pscomp(const void *a, const void *b) { const KINFO *ka, *kb; ka = a; kb = b; /* SORTCPU and SORTMEM are sorted in descending order. */ if (sortby == SORTCPU) DIFF_RETURN(kb, ka, ki_pcpu); if (sortby == SORTMEM) DIFF_RETURN(kb, ka, ki_memsize); /* * TTY's are sorted in ascending order, except that all NODEV * processes come before all processes with a device. */ if (ka->ki_p->ki_tdev != kb->ki_p->ki_tdev) { if (ka->ki_p->ki_tdev == NODEV) return (-1); if (kb->ki_p->ki_tdev == NODEV) return (1); DIFF_RETURN(ka, kb, ki_p->ki_tdev); } /* PID's and TID's (threads) are sorted in ascending order. */ DIFF_RETURN(ka, kb, ki_p->ki_pid); DIFF_RETURN(ka, kb, ki_p->ki_tid); return (0); } #undef DIFF_RETURN /* * ICK (all for getopt), would rather hide the ugliness * here than taint the main code. * * ps foo -> ps -foo * ps 34 -> ps -p34 * * The old convention that 't' with no trailing tty arg means the users * tty, is only supported if argv[1] doesn't begin with a '-'. This same * feature is available with the option 'T', which takes no argument. */ static char * kludge_oldps_options(const char *optlist, char *origval, const char *nextarg) { size_t len; char *argp, *cp, *newopts, *ns, *optp, *pidp; /* * See if the original value includes any option which takes an * argument (and will thus use up the remainder of the string). */ argp = NULL; if (optlist != NULL) { for (cp = origval; *cp != '\0'; cp++) { optp = strchr(optlist, *cp); if ((optp != NULL) && *(optp + 1) == ':') { argp = cp; break; } } } if (argp != NULL && *origval == '-') return (origval); /* * if last letter is a 't' flag with no argument (in the context * of the oldps options -- option string NOT starting with a '-' -- * then convert to 'T' (meaning *this* terminal, i.e. ttyname(0)). * * However, if a flag accepting a string argument is found earlier * in the option string (including a possible `t' flag), then the * remainder of the string must be the argument to that flag; so * do not modify that argument. Note that a trailing `t' would * cause argp to be set, if argp was not already set by some * earlier option. */ len = strlen(origval); cp = origval + len - 1; pidp = NULL; if (*cp == 't' && *origval != '-' && cp == argp) { if (nextarg == NULL || *nextarg == '-' || isdigitch(*nextarg)) *cp = 'T'; } else if (argp == NULL) { /* * The original value did not include any option which takes * an argument (and that would include `p' and `t'), so check * the value for trailing number, or comma-separated list of * numbers, which we will treat as a pid request. */ if (isdigitch(*cp)) { while (cp >= origval && (*cp == ',' || isdigitch(*cp))) --cp; pidp = cp + 1; } } /* * If nothing needs to be added to the string, then return * the "original" (although possibly modified) value. */ if (*origval == '-' && pidp == NULL) return (origval); /* * Create a copy of the string to add '-' and/or 'p' to the * original value. */ if ((newopts = ns = malloc(len + 3)) == NULL) xo_errx(1, "malloc failed"); if (*origval != '-') *ns++ = '-'; /* add option flag */ if (pidp == NULL) strcpy(ns, origval); else { /* * Copy everything before the pid string, add the `p', * and then copy the pid string. */ len = pidp - origval; memcpy(ns, origval, len); ns += len; *ns++ = 'p'; strcpy(ns, pidp); } return (newopts); } static void pidmax_init(void) { size_t intsize; intsize = sizeof(pid_max); if (sysctlbyname("kern.pid_max", &pid_max, &intsize, NULL, 0) < 0) { xo_warn("unable to read kern.pid_max"); pid_max = 99999; } } static void usage(void) { #define SINGLE_OPTS "[-aCcde" OPT_LAZY_f "HhjlmrSTuvwXxZ]" (void)xo_error("%s\n%s\n%s\n%s\n", "usage: ps " SINGLE_OPTS " [-O fmt | -o fmt] [-G gid[,gid...]]", " [-J jid[,jid...]] [-M core] [-N system]", " [-p pid[,pid...]] [-t tty[,tty...]] [-U user[,user...]]", " ps [-L]"); exit(1); } Index: projects/clang-trunk/contrib/diff/src/context.c =================================================================== --- projects/clang-trunk/contrib/diff/src/context.c (revision 283870) +++ projects/clang-trunk/contrib/diff/src/context.c (revision 283871) @@ -1,478 +1,478 @@ /* Context-format output routines for GNU DIFF. Copyright (C) 1988, 1989, 1991, 1992, 1993, 1994, 1995, 1998, 2001, 2002, 2004 Free Software Foundation, Inc. This file is part of GNU DIFF. GNU DIFF is free software; you can redistribute it and/or modify it under the terms of the GNU General Public License as published by the Free Software Foundation; either version 2, or (at your option) any later version. GNU DIFF is distributed in the hope that it will be useful, but WITHOUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License for more details. You should have received a copy of the GNU General Public License along with this program; see the file COPYING. If not, write to the Free Software Foundation, 59 Temple Place - Suite 330, Boston, MA 02111-1307, USA. */ #include "diff.h" #include #ifdef ST_MTIM_NSEC # define TIMESPEC_NS(timespec) ((timespec).ST_MTIM_NSEC) #else # define TIMESPEC_NS(timespec) 0 #endif size_t nstrftime (char *, size_t, char const *, struct tm const *, int, long); static char const *find_function (char const * const *, lin); static struct change *find_hunk (struct change *); static void mark_ignorable (struct change *); static void pr_context_hunk (struct change *); static void pr_unidiff_hunk (struct change *); /* Last place find_function started searching from. */ static lin find_function_last_search; /* The value find_function returned when it started searching there. */ static lin find_function_last_match; /* Print a label for a context diff, with a file name and date or a label. */ static void print_context_label (char const *mark, struct file_data *inf, char const *label) { if (label) fprintf (outfile, "%s %s\n", mark, label); else { char buf[MAX (INT_STRLEN_BOUND (int) + 32, INT_STRLEN_BOUND (time_t) + 11)]; struct tm const *tm = localtime (&inf->stat.st_mtime); long nsec = TIMESPEC_NS (inf->stat.st_mtim); if (! (tm && nstrftime (buf, sizeof buf, time_format, tm, 0, nsec))) { time_t sec = inf->stat.st_mtime; verify (info_preserved, sizeof inf->stat.st_mtime <= sizeof sec); - sprintf (buf, "%jd.%.9d", (intmax_t)sec, nsec); + sprintf (buf, "%jd.%.9ld", (intmax_t)sec, nsec); } fprintf (outfile, "%s %s\t%s\n", mark, inf->name, buf); } } /* Print a header for a context diff, with the file names and dates. */ void print_context_header (struct file_data inf[], bool unidiff) { if (unidiff) { print_context_label ("---", &inf[0], file_label[0]); print_context_label ("+++", &inf[1], file_label[1]); } else { print_context_label ("***", &inf[0], file_label[0]); print_context_label ("---", &inf[1], file_label[1]); } } /* Print an edit script in context format. */ void print_context_script (struct change *script, bool unidiff) { if (ignore_blank_lines || ignore_regexp.fastmap) mark_ignorable (script); else { struct change *e; for (e = script; e; e = e->link) e->ignore = false; } find_function_last_search = - files[0].prefix_lines; find_function_last_match = LIN_MAX; if (unidiff) print_script (script, find_hunk, pr_unidiff_hunk); else print_script (script, find_hunk, pr_context_hunk); } /* Print a pair of line numbers with a comma, translated for file FILE. If the second number is not greater, use the first in place of it. Args A and B are internal line numbers. We print the translated (real) line numbers. */ static void print_context_number_range (struct file_data const *file, lin a, lin b) { long int trans_a, trans_b; translate_range (file, a, b, &trans_a, &trans_b); /* We can have B <= A in the case of a range of no lines. In this case, we should print the line number before the range, which is B. POSIX 1003.1-2001 requires two line numbers separated by a comma even if the line numbers are the same. However, this does not match existing practice and is surely an error in the specification. */ if (trans_b <= trans_a) fprintf (outfile, "%ld", trans_b); else fprintf (outfile, "%ld,%ld", trans_a, trans_b); } /* Print FUNCTION in a context header. */ static void print_context_function (FILE *out, char const *function) { int i; putc (' ', out); for (i = 0; i < 40 && function[i] != '\n'; i++) continue; fwrite (function, sizeof (char), i, out); } /* Print a portion of an edit script in context format. HUNK is the beginning of the portion to be printed. The end is marked by a `link' that has been nulled out. Prints out lines from both files, and precedes each line with the appropriate flag-character. */ static void pr_context_hunk (struct change *hunk) { lin first0, last0, first1, last1, i; char const *prefix; char const *function; FILE *out; /* Determine range of line numbers involved in each file. */ enum changes changes = analyze_hunk (hunk, &first0, &last0, &first1, &last1); if (! changes) return; /* Include a context's width before and after. */ i = - files[0].prefix_lines; first0 = MAX (first0 - context, i); first1 = MAX (first1 - context, i); if (last0 < files[0].valid_lines - context) last0 += context; else last0 = files[0].valid_lines - 1; if (last1 < files[1].valid_lines - context) last1 += context; else last1 = files[1].valid_lines - 1; /* If desired, find the preceding function definition line in file 0. */ function = 0; if (function_regexp.fastmap) function = find_function (files[0].linbuf, first0); begin_output (); out = outfile; fprintf (out, "***************"); if (function) print_context_function (out, function); fprintf (out, "\n*** "); print_context_number_range (&files[0], first0, last0); fprintf (out, " ****\n"); if (changes & OLD) { struct change *next = hunk; for (i = first0; i <= last0; i++) { /* Skip past changes that apply (in file 0) only to lines before line I. */ while (next && next->line0 + next->deleted <= i) next = next->link; /* Compute the marking for line I. */ prefix = " "; if (next && next->line0 <= i) /* The change NEXT covers this line. If lines were inserted here in file 1, this is "changed". Otherwise it is "deleted". */ prefix = (next->inserted > 0 ? "!" : "-"); print_1_line (prefix, &files[0].linbuf[i]); } } fprintf (out, "--- "); print_context_number_range (&files[1], first1, last1); fprintf (out, " ----\n"); if (changes & NEW) { struct change *next = hunk; for (i = first1; i <= last1; i++) { /* Skip past changes that apply (in file 1) only to lines before line I. */ while (next && next->line1 + next->inserted <= i) next = next->link; /* Compute the marking for line I. */ prefix = " "; if (next && next->line1 <= i) /* The change NEXT covers this line. If lines were deleted here in file 0, this is "changed". Otherwise it is "inserted". */ prefix = (next->deleted > 0 ? "!" : "+"); print_1_line (prefix, &files[1].linbuf[i]); } } } /* Print a pair of line numbers with a comma, translated for file FILE. If the second number is smaller, use the first in place of it. If the numbers are equal, print just one number. Args A and B are internal line numbers. We print the translated (real) line numbers. */ static void print_unidiff_number_range (struct file_data const *file, lin a, lin b) { long int trans_a, trans_b; translate_range (file, a, b, &trans_a, &trans_b); /* We can have B < A in the case of a range of no lines. In this case, we print the line number before the range, which is B. It would be more logical to print A, but 'patch' expects B in order to detect diffs against empty files. */ if (trans_b <= trans_a) fprintf (outfile, trans_b < trans_a ? "%ld,0" : "%ld", trans_b); else fprintf (outfile, "%ld,%ld", trans_a, trans_b - trans_a + 1); } /* Print a portion of an edit script in unidiff format. HUNK is the beginning of the portion to be printed. The end is marked by a `link' that has been nulled out. Prints out lines from both files, and precedes each line with the appropriate flag-character. */ static void pr_unidiff_hunk (struct change *hunk) { lin first0, last0, first1, last1; lin i, j, k; struct change *next; char const *function; FILE *out; /* Determine range of line numbers involved in each file. */ if (! analyze_hunk (hunk, &first0, &last0, &first1, &last1)) return; /* Include a context's width before and after. */ i = - files[0].prefix_lines; first0 = MAX (first0 - context, i); first1 = MAX (first1 - context, i); if (last0 < files[0].valid_lines - context) last0 += context; else last0 = files[0].valid_lines - 1; if (last1 < files[1].valid_lines - context) last1 += context; else last1 = files[1].valid_lines - 1; /* If desired, find the preceding function definition line in file 0. */ function = 0; if (function_regexp.fastmap) function = find_function (files[0].linbuf, first0); begin_output (); out = outfile; fprintf (out, "@@ -"); print_unidiff_number_range (&files[0], first0, last0); fprintf (out, " +"); print_unidiff_number_range (&files[1], first1, last1); fprintf (out, " @@"); if (function) print_context_function (out, function); putc ('\n', out); next = hunk; i = first0; j = first1; while (i <= last0 || j <= last1) { /* If the line isn't a difference, output the context from file 0. */ if (!next || i < next->line0) { putc (initial_tab ? '\t' : ' ', out); print_1_line (0, &files[0].linbuf[i++]); j++; } else { /* For each difference, first output the deleted part. */ k = next->deleted; while (k--) { putc ('-', out); if (initial_tab) putc ('\t', out); print_1_line (0, &files[0].linbuf[i++]); } /* Then output the inserted part. */ k = next->inserted; while (k--) { putc ('+', out); if (initial_tab) putc ('\t', out); print_1_line (0, &files[1].linbuf[j++]); } /* We're done with this hunk, so on to the next! */ next = next->link; } } } /* Scan a (forward-ordered) edit script for the first place that more than 2*CONTEXT unchanged lines appear, and return a pointer to the `struct change' for the last change before those lines. */ static struct change * find_hunk (struct change *start) { struct change *prev; lin top0, top1; lin thresh; /* Threshold distance is 2 * CONTEXT + 1 between two non-ignorable changes, but only CONTEXT if one is ignorable. Watch out for integer overflow, though. */ lin non_ignorable_threshold = (LIN_MAX - 1) / 2 < context ? LIN_MAX : 2 * context + 1; lin ignorable_threshold = context; do { /* Compute number of first line in each file beyond this changed. */ top0 = start->line0 + start->deleted; top1 = start->line1 + start->inserted; prev = start; start = start->link; thresh = (prev->ignore || (start && start->ignore) ? ignorable_threshold : non_ignorable_threshold); /* It is not supposed to matter which file we check in the end-test. If it would matter, crash. */ if (start && start->line0 - top0 != start->line1 - top1) abort (); } while (start /* Keep going if less than THRESH lines elapse before the affected line. */ && start->line0 - top0 < thresh); return prev; } /* Set the `ignore' flag properly in each change in SCRIPT. It should be 1 if all the lines inserted or deleted in that change are ignorable lines. */ static void mark_ignorable (struct change *script) { while (script) { struct change *next = script->link; lin first0, last0, first1, last1; /* Turn this change into a hunk: detach it from the others. */ script->link = 0; /* Determine whether this change is ignorable. */ script->ignore = ! analyze_hunk (script, &first0, &last0, &first1, &last1); /* Reconnect the chain as before. */ script->link = next; /* Advance to the following change. */ script = next; } } /* Find the last function-header line in LINBUF prior to line number LINENUM. This is a line containing a match for the regexp in `function_regexp'. Return the address of the text, or 0 if no function-header is found. */ static char const * find_function (char const * const *linbuf, lin linenum) { lin i = linenum; lin last = find_function_last_search; find_function_last_search = i; while (last <= --i) { /* See if this line is what we want. */ char const *line = linbuf[i]; size_t linelen = linbuf[i + 1] - line - 1; /* FIXME: re_search's size args should be size_t, not int. */ int len = MIN (linelen, INT_MAX); if (0 <= re_search (&function_regexp, line, len, 0, len, 0)) { find_function_last_match = i; return line; } } /* If we search back to where we started searching the previous time, find the line we found last time. */ if (find_function_last_match != LIN_MAX) return linbuf[find_function_last_match]; return 0; } Index: projects/clang-trunk/contrib/gcc/config/arm/lib1funcs.asm =================================================================== --- projects/clang-trunk/contrib/gcc/config/arm/lib1funcs.asm (revision 283870) +++ projects/clang-trunk/contrib/gcc/config/arm/lib1funcs.asm (revision 283871) @@ -1,1307 +1,1307 @@ @ libgcc routines for ARM cpu. @ Division routines, written by Richard Earnshaw, (rearnsha@armltd.co.uk) /* Copyright 1995, 1996, 1998, 1999, 2000, 2003, 2004, 2005 Free Software Foundation, Inc. This file is free software; you can redistribute it and/or modify it under the terms of the GNU General Public License as published by the Free Software Foundation; either version 2, or (at your option) any later version. In addition to the permissions in the GNU General Public License, the Free Software Foundation gives you unlimited permission to link the compiled version of this file into combinations with other programs, and to distribute those combinations without any restriction coming from the use of this file. (The General Public License restrictions do apply in other respects; for example, they cover modification of the file, and distribution when not linked into a combine executable.) This file is distributed in the hope that it will be useful, but WITHOUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License for more details. You should have received a copy of the GNU General Public License along with this program; see the file COPYING. If not, write to the Free Software Foundation, 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301, USA. */ /* ------------------------------------------------------------------------ */ /* We need to know what prefix to add to function names. */ #ifndef __USER_LABEL_PREFIX__ #error __USER_LABEL_PREFIX__ not defined #endif /* ANSI concatenation macros. */ #define CONCAT1(a, b) CONCAT2(a, b) #define CONCAT2(a, b) a ## b /* Use the right prefix for global labels. */ #define SYM(x) CONCAT1 (__USER_LABEL_PREFIX__, x) #ifdef __ELF__ #ifdef __thumb__ #define __PLT__ /* Not supported in Thumb assembler (for now). */ #else #define __PLT__ (PLT) #endif #define TYPE(x) .type SYM(x),function #define SIZE(x) .size SYM(x), . - SYM(x) #define LSYM(x) .x #else #define __PLT__ #define TYPE(x) #define SIZE(x) #define LSYM(x) x #endif /* Function end macros. Variants for interworking. */ @ This selects the minimum architecture level required. #define __ARM_ARCH__ 3 #if defined(__ARM_ARCH_3M__) || defined(__ARM_ARCH_4__) \ || defined(__ARM_ARCH_4T__) /* We use __ARM_ARCH__ set to 4 here, but in reality it's any processor with long multiply instructions. That includes v3M. */ # undef __ARM_ARCH__ # define __ARM_ARCH__ 4 #endif #if defined(__ARM_ARCH_5__) || defined(__ARM_ARCH_5T__) \ || defined(__ARM_ARCH_5E__) || defined(__ARM_ARCH_5TE__) \ || defined(__ARM_ARCH_5TEJ__) # undef __ARM_ARCH__ # define __ARM_ARCH__ 5 #endif #if defined(__ARM_ARCH_6__) || defined(__ARM_ARCH_6J__) \ || defined(__ARM_ARCH_6K__) || defined(__ARM_ARCH_6Z__) \ || defined(__ARM_ARCH_6ZK__) # undef __ARM_ARCH__ # define __ARM_ARCH__ 6 #endif #ifndef __ARM_ARCH__ #error Unable to determine architecture. #endif /* How to return from a function call depends on the architecture variant. */ #if (__ARM_ARCH__ > 4) || defined(__ARM_ARCH_4T__) # define RET bx lr # define RETc(x) bx##x lr /* Special precautions for interworking on armv4t. */ # if (__ARM_ARCH__ == 4) /* Always use bx, not ldr pc. */ # if (defined(__thumb__) || defined(__THUMB_INTERWORK__)) # define __INTERWORKING__ # endif /* __THUMB__ || __THUMB_INTERWORK__ */ /* Include thumb stub before arm mode code. */ # if defined(__thumb__) && !defined(__THUMB_INTERWORK__) # define __INTERWORKING_STUBS__ # endif /* __thumb__ && !__THUMB_INTERWORK__ */ #endif /* __ARM_ARCH == 4 */ #else # define RET mov pc, lr # define RETc(x) mov##x pc, lr #endif .macro cfi_pop advance, reg, cfa_offset #ifdef __ELF__ .pushsection .debug_frame .byte 0x4 /* DW_CFA_advance_loc4 */ .4byte \advance .byte (0xc0 | \reg) /* DW_CFA_restore */ .byte 0xe /* DW_CFA_def_cfa_offset */ .uleb128 \cfa_offset .popsection #endif .endm .macro cfi_push advance, reg, offset, cfa_offset #ifdef __ELF__ .pushsection .debug_frame .byte 0x4 /* DW_CFA_advance_loc4 */ .4byte \advance .byte (0x80 | \reg) /* DW_CFA_offset */ .uleb128 (\offset / -4) .byte 0xe /* DW_CFA_def_cfa_offset */ .uleb128 \cfa_offset .popsection #endif .endm .macro cfi_start start_label, end_label #ifdef __ELF__ .pushsection .debug_frame LSYM(Lstart_frame): .4byte LSYM(Lend_cie) - LSYM(Lstart_cie) @ Length of CIE LSYM(Lstart_cie): .4byte 0xffffffff @ CIE Identifier Tag .byte 0x1 @ CIE Version .ascii "\0" @ CIE Augmentation .uleb128 0x1 @ CIE Code Alignment Factor .sleb128 -4 @ CIE Data Alignment Factor .byte 0xe @ CIE RA Column .byte 0xc @ DW_CFA_def_cfa .uleb128 0xd .uleb128 0x0 .align 2 LSYM(Lend_cie): .4byte LSYM(Lend_fde)-LSYM(Lstart_fde) @ FDE Length LSYM(Lstart_fde): .4byte LSYM(Lstart_frame) @ FDE CIE offset .4byte \start_label @ FDE initial location .4byte \end_label-\start_label @ FDE address range .popsection #endif .endm .macro cfi_end end_label #ifdef __ELF__ .pushsection .debug_frame .align 2 LSYM(Lend_fde): .popsection \end_label: #endif .endm /* Don't pass dirn, it's there just to get token pasting right. */ .macro RETLDM regs=, cond=, unwind=, dirn=ia #if defined (__INTERWORKING__) .ifc "\regs","" ldr\cond lr, [sp], #8 .else ldm\cond\dirn sp!, {\regs, lr} .endif .ifnc "\unwind", "" /* Mark LR as restored. */ 97: cfi_pop 97b - \unwind, 0xe, 0x0 .endif bx\cond lr #else .ifc "\regs","" ldr\cond pc, [sp], #8 .else ldm\cond\dirn sp!, {\regs, pc} .endif #endif .endm .macro ARM_LDIV0 name str lr, [sp, #-8]! 98: cfi_push 98b - __\name, 0xe, -0x8, 0x8 bl SYM (__div0) __PLT__ mov r0, #0 @ About as wrong as it could be. RETLDM unwind=98b .endm .macro THUMB_LDIV0 name push { r1, lr } 98: cfi_push 98b - __\name, 0xe, -0x4, 0x8 bl SYM (__div0) mov r0, #0 @ About as wrong as it could be. #if defined (__INTERWORKING__) pop { r1, r2 } bx r2 #else pop { r1, pc } #endif .endm .macro FUNC_END name SIZE (__\name) .endm .macro DIV_FUNC_END name cfi_start __\name, LSYM(Lend_div0) LSYM(Ldiv0): #ifdef __thumb__ THUMB_LDIV0 \name #else ARM_LDIV0 \name #endif cfi_end LSYM(Lend_div0) FUNC_END \name .endm .macro THUMB_FUNC_START name .globl SYM (\name) TYPE (\name) .thumb_func SYM (\name): .endm /* Function start macros. Variants for ARM and Thumb. */ #ifdef __thumb__ #define THUMB_FUNC .thumb_func -#define THUMB_CODE .force_thumb +#define THUMB_CODE .thumb #else #define THUMB_FUNC #define THUMB_CODE #endif .macro FUNC_START name .text .globl SYM (__\name) TYPE (__\name) .align 0 THUMB_CODE THUMB_FUNC SYM (__\name): .endm /* Special function that will always be coded in ARM assembly, even if in Thumb-only compilation. */ #if defined(__INTERWORKING_STUBS__) .macro ARM_FUNC_START name FUNC_START \name bx pc nop .arm /* A hook to tell gdb that we've switched to ARM mode. Also used to call directly from other local arm routines. */ _L__\name: .endm #define EQUIV .thumb_set /* Branch directly to a function declared with ARM_FUNC_START. Must be called in arm mode. */ .macro ARM_CALL name bl _L__\name .endm #else .macro ARM_FUNC_START name .text .globl SYM (__\name) TYPE (__\name) .align 0 .arm SYM (__\name): .endm #define EQUIV .set .macro ARM_CALL name bl __\name .endm #endif .macro FUNC_ALIAS new old .globl SYM (__\new) #if defined (__thumb__) .thumb_set SYM (__\new), SYM (__\old) #else .set SYM (__\new), SYM (__\old) #endif .endm .macro ARM_FUNC_ALIAS new old .globl SYM (__\new) EQUIV SYM (__\new), SYM (__\old) #if defined(__INTERWORKING_STUBS__) .set SYM (_L__\new), SYM (_L__\old) #endif .endm #ifdef __thumb__ /* Register aliases. */ work .req r4 @ XXXX is this safe ? dividend .req r0 divisor .req r1 overdone .req r2 result .req r2 curbit .req r3 #endif #if 0 ip .req r12 sp .req r13 lr .req r14 pc .req r15 #endif /* ------------------------------------------------------------------------ */ /* Bodies of the division and modulo routines. */ /* ------------------------------------------------------------------------ */ .macro ARM_DIV_BODY dividend, divisor, result, curbit #if __ARM_ARCH__ >= 5 && ! defined (__OPTIMIZE_SIZE__) clz \curbit, \dividend clz \result, \divisor sub \curbit, \result, \curbit rsbs \curbit, \curbit, #31 addne \curbit, \curbit, \curbit, lsl #1 mov \result, #0 addne pc, pc, \curbit, lsl #2 nop .set shift, 32 .rept 32 .set shift, shift - 1 cmp \dividend, \divisor, lsl #shift adc \result, \result, \result subcs \dividend, \dividend, \divisor, lsl #shift .endr #else /* __ARM_ARCH__ < 5 || defined (__OPTIMIZE_SIZE__) */ #if __ARM_ARCH__ >= 5 clz \curbit, \divisor clz \result, \dividend sub \result, \curbit, \result mov \curbit, #1 mov \divisor, \divisor, lsl \result mov \curbit, \curbit, lsl \result mov \result, #0 #else /* __ARM_ARCH__ < 5 */ @ Initially shift the divisor left 3 bits if possible, @ set curbit accordingly. This allows for curbit to be located @ at the left end of each 4 bit nibbles in the division loop @ to save one loop in most cases. tst \divisor, #0xe0000000 moveq \divisor, \divisor, lsl #3 moveq \curbit, #8 movne \curbit, #1 @ Unless the divisor is very big, shift it up in multiples of @ four bits, since this is the amount of unwinding in the main @ division loop. Continue shifting until the divisor is @ larger than the dividend. 1: cmp \divisor, #0x10000000 cmplo \divisor, \dividend movlo \divisor, \divisor, lsl #4 movlo \curbit, \curbit, lsl #4 blo 1b @ For very big divisors, we must shift it a bit at a time, or @ we will be in danger of overflowing. 1: cmp \divisor, #0x80000000 cmplo \divisor, \dividend movlo \divisor, \divisor, lsl #1 movlo \curbit, \curbit, lsl #1 blo 1b mov \result, #0 #endif /* __ARM_ARCH__ < 5 */ @ Division loop 1: cmp \dividend, \divisor subhs \dividend, \dividend, \divisor orrhs \result, \result, \curbit cmp \dividend, \divisor, lsr #1 subhs \dividend, \dividend, \divisor, lsr #1 orrhs \result, \result, \curbit, lsr #1 cmp \dividend, \divisor, lsr #2 subhs \dividend, \dividend, \divisor, lsr #2 orrhs \result, \result, \curbit, lsr #2 cmp \dividend, \divisor, lsr #3 subhs \dividend, \dividend, \divisor, lsr #3 orrhs \result, \result, \curbit, lsr #3 cmp \dividend, #0 @ Early termination? movnes \curbit, \curbit, lsr #4 @ No, any more bits to do? movne \divisor, \divisor, lsr #4 bne 1b #endif /* __ARM_ARCH__ < 5 || defined (__OPTIMIZE_SIZE__) */ .endm /* ------------------------------------------------------------------------ */ .macro ARM_DIV2_ORDER divisor, order #if __ARM_ARCH__ >= 5 clz \order, \divisor rsb \order, \order, #31 #else cmp \divisor, #(1 << 16) movhs \divisor, \divisor, lsr #16 movhs \order, #16 movlo \order, #0 cmp \divisor, #(1 << 8) movhs \divisor, \divisor, lsr #8 addhs \order, \order, #8 cmp \divisor, #(1 << 4) movhs \divisor, \divisor, lsr #4 addhs \order, \order, #4 cmp \divisor, #(1 << 2) addhi \order, \order, #3 addls \order, \order, \divisor, lsr #1 #endif .endm /* ------------------------------------------------------------------------ */ .macro ARM_MOD_BODY dividend, divisor, order, spare #if __ARM_ARCH__ >= 5 && ! defined (__OPTIMIZE_SIZE__) clz \order, \divisor clz \spare, \dividend sub \order, \order, \spare rsbs \order, \order, #31 addne pc, pc, \order, lsl #3 nop .set shift, 32 .rept 32 .set shift, shift - 1 cmp \dividend, \divisor, lsl #shift subcs \dividend, \dividend, \divisor, lsl #shift .endr #else /* __ARM_ARCH__ < 5 || defined (__OPTIMIZE_SIZE__) */ #if __ARM_ARCH__ >= 5 clz \order, \divisor clz \spare, \dividend sub \order, \order, \spare mov \divisor, \divisor, lsl \order #else /* __ARM_ARCH__ < 5 */ mov \order, #0 @ Unless the divisor is very big, shift it up in multiples of @ four bits, since this is the amount of unwinding in the main @ division loop. Continue shifting until the divisor is @ larger than the dividend. 1: cmp \divisor, #0x10000000 cmplo \divisor, \dividend movlo \divisor, \divisor, lsl #4 addlo \order, \order, #4 blo 1b @ For very big divisors, we must shift it a bit at a time, or @ we will be in danger of overflowing. 1: cmp \divisor, #0x80000000 cmplo \divisor, \dividend movlo \divisor, \divisor, lsl #1 addlo \order, \order, #1 blo 1b #endif /* __ARM_ARCH__ < 5 */ @ Perform all needed substractions to keep only the reminder. @ Do comparisons in batch of 4 first. subs \order, \order, #3 @ yes, 3 is intended here blt 2f 1: cmp \dividend, \divisor subhs \dividend, \dividend, \divisor cmp \dividend, \divisor, lsr #1 subhs \dividend, \dividend, \divisor, lsr #1 cmp \dividend, \divisor, lsr #2 subhs \dividend, \dividend, \divisor, lsr #2 cmp \dividend, \divisor, lsr #3 subhs \dividend, \dividend, \divisor, lsr #3 cmp \dividend, #1 mov \divisor, \divisor, lsr #4 subges \order, \order, #4 bge 1b tst \order, #3 teqne \dividend, #0 beq 5f @ Either 1, 2 or 3 comparison/substractions are left. 2: cmn \order, #2 blt 4f beq 3f cmp \dividend, \divisor subhs \dividend, \dividend, \divisor mov \divisor, \divisor, lsr #1 3: cmp \dividend, \divisor subhs \dividend, \dividend, \divisor mov \divisor, \divisor, lsr #1 4: cmp \dividend, \divisor subhs \dividend, \dividend, \divisor 5: #endif /* __ARM_ARCH__ < 5 || defined (__OPTIMIZE_SIZE__) */ .endm /* ------------------------------------------------------------------------ */ .macro THUMB_DIV_MOD_BODY modulo @ Load the constant 0x10000000 into our work register. mov work, #1 lsl work, #28 LSYM(Loop1): @ Unless the divisor is very big, shift it up in multiples of @ four bits, since this is the amount of unwinding in the main @ division loop. Continue shifting until the divisor is @ larger than the dividend. cmp divisor, work bhs LSYM(Lbignum) cmp divisor, dividend bhs LSYM(Lbignum) lsl divisor, #4 lsl curbit, #4 b LSYM(Loop1) LSYM(Lbignum): @ Set work to 0x80000000 lsl work, #3 LSYM(Loop2): @ For very big divisors, we must shift it a bit at a time, or @ we will be in danger of overflowing. cmp divisor, work bhs LSYM(Loop3) cmp divisor, dividend bhs LSYM(Loop3) lsl divisor, #1 lsl curbit, #1 b LSYM(Loop2) LSYM(Loop3): @ Test for possible subtractions ... .if \modulo @ ... On the final pass, this may subtract too much from the dividend, @ so keep track of which subtractions are done, we can fix them up @ afterwards. mov overdone, #0 cmp dividend, divisor blo LSYM(Lover1) sub dividend, dividend, divisor LSYM(Lover1): lsr work, divisor, #1 cmp dividend, work blo LSYM(Lover2) sub dividend, dividend, work mov ip, curbit mov work, #1 ror curbit, work orr overdone, curbit mov curbit, ip LSYM(Lover2): lsr work, divisor, #2 cmp dividend, work blo LSYM(Lover3) sub dividend, dividend, work mov ip, curbit mov work, #2 ror curbit, work orr overdone, curbit mov curbit, ip LSYM(Lover3): lsr work, divisor, #3 cmp dividend, work blo LSYM(Lover4) sub dividend, dividend, work mov ip, curbit mov work, #3 ror curbit, work orr overdone, curbit mov curbit, ip LSYM(Lover4): mov ip, curbit .else @ ... and note which bits are done in the result. On the final pass, @ this may subtract too much from the dividend, but the result will be ok, @ since the "bit" will have been shifted out at the bottom. cmp dividend, divisor blo LSYM(Lover1) sub dividend, dividend, divisor orr result, result, curbit LSYM(Lover1): lsr work, divisor, #1 cmp dividend, work blo LSYM(Lover2) sub dividend, dividend, work lsr work, curbit, #1 orr result, work LSYM(Lover2): lsr work, divisor, #2 cmp dividend, work blo LSYM(Lover3) sub dividend, dividend, work lsr work, curbit, #2 orr result, work LSYM(Lover3): lsr work, divisor, #3 cmp dividend, work blo LSYM(Lover4) sub dividend, dividend, work lsr work, curbit, #3 orr result, work LSYM(Lover4): .endif cmp dividend, #0 @ Early termination? beq LSYM(Lover5) lsr curbit, #4 @ No, any more bits to do? beq LSYM(Lover5) lsr divisor, #4 b LSYM(Loop3) LSYM(Lover5): .if \modulo @ Any subtractions that we should not have done will be recorded in @ the top three bits of "overdone". Exactly which were not needed @ are governed by the position of the bit, stored in ip. mov work, #0xe lsl work, #28 and overdone, work beq LSYM(Lgot_result) @ If we terminated early, because dividend became zero, then the @ bit in ip will not be in the bottom nibble, and we should not @ perform the additions below. We must test for this though @ (rather relying upon the TSTs to prevent the additions) since @ the bit in ip could be in the top two bits which might then match @ with one of the smaller RORs. mov curbit, ip mov work, #0x7 tst curbit, work beq LSYM(Lgot_result) mov curbit, ip mov work, #3 ror curbit, work tst overdone, curbit beq LSYM(Lover6) lsr work, divisor, #3 add dividend, work LSYM(Lover6): mov curbit, ip mov work, #2 ror curbit, work tst overdone, curbit beq LSYM(Lover7) lsr work, divisor, #2 add dividend, work LSYM(Lover7): mov curbit, ip mov work, #1 ror curbit, work tst overdone, curbit beq LSYM(Lgot_result) lsr work, divisor, #1 add dividend, work .endif LSYM(Lgot_result): .endm /* ------------------------------------------------------------------------ */ /* Start of the Real Functions */ /* ------------------------------------------------------------------------ */ #ifdef L_udivsi3 FUNC_START udivsi3 FUNC_ALIAS aeabi_uidiv udivsi3 #ifdef __thumb__ cmp divisor, #0 beq LSYM(Ldiv0) mov curbit, #1 mov result, #0 push { work } cmp dividend, divisor blo LSYM(Lgot_result) THUMB_DIV_MOD_BODY 0 mov r0, result pop { work } RET #else /* ARM version. */ subs r2, r1, #1 RETc(eq) bcc LSYM(Ldiv0) cmp r0, r1 bls 11f tst r1, r2 beq 12f ARM_DIV_BODY r0, r1, r2, r3 mov r0, r2 RET 11: moveq r0, #1 movne r0, #0 RET 12: ARM_DIV2_ORDER r1, r2 mov r0, r0, lsr r2 RET #endif /* ARM version */ DIV_FUNC_END udivsi3 FUNC_START aeabi_uidivmod #ifdef __thumb__ push {r0, r1, lr} bl SYM(__udivsi3) POP {r1, r2, r3} mul r2, r0 sub r1, r1, r2 bx r3 #else stmfd sp!, { r0, r1, lr } bl SYM(__udivsi3) ldmfd sp!, { r1, r2, lr } mul r3, r2, r0 sub r1, r1, r3 RET #endif FUNC_END aeabi_uidivmod #endif /* L_udivsi3 */ /* ------------------------------------------------------------------------ */ #ifdef L_umodsi3 FUNC_START umodsi3 #ifdef __thumb__ cmp divisor, #0 beq LSYM(Ldiv0) mov curbit, #1 cmp dividend, divisor bhs LSYM(Lover10) RET LSYM(Lover10): push { work } THUMB_DIV_MOD_BODY 1 pop { work } RET #else /* ARM version. */ subs r2, r1, #1 @ compare divisor with 1 bcc LSYM(Ldiv0) cmpne r0, r1 @ compare dividend with divisor moveq r0, #0 tsthi r1, r2 @ see if divisor is power of 2 andeq r0, r0, r2 RETc(ls) ARM_MOD_BODY r0, r1, r2, r3 RET #endif /* ARM version. */ DIV_FUNC_END umodsi3 #endif /* L_umodsi3 */ /* ------------------------------------------------------------------------ */ #ifdef L_divsi3 FUNC_START divsi3 FUNC_ALIAS aeabi_idiv divsi3 #ifdef __thumb__ cmp divisor, #0 beq LSYM(Ldiv0) push { work } mov work, dividend eor work, divisor @ Save the sign of the result. mov ip, work mov curbit, #1 mov result, #0 cmp divisor, #0 bpl LSYM(Lover10) neg divisor, divisor @ Loops below use unsigned. LSYM(Lover10): cmp dividend, #0 bpl LSYM(Lover11) neg dividend, dividend LSYM(Lover11): cmp dividend, divisor blo LSYM(Lgot_result) THUMB_DIV_MOD_BODY 0 mov r0, result mov work, ip cmp work, #0 bpl LSYM(Lover12) neg r0, r0 LSYM(Lover12): pop { work } RET #else /* ARM version. */ cmp r1, #0 eor ip, r0, r1 @ save the sign of the result. beq LSYM(Ldiv0) rsbmi r1, r1, #0 @ loops below use unsigned. subs r2, r1, #1 @ division by 1 or -1 ? beq 10f movs r3, r0 rsbmi r3, r0, #0 @ positive dividend value cmp r3, r1 bls 11f tst r1, r2 @ divisor is power of 2 ? beq 12f ARM_DIV_BODY r3, r1, r0, r2 cmp ip, #0 rsbmi r0, r0, #0 RET 10: teq ip, r0 @ same sign ? rsbmi r0, r0, #0 RET 11: movlo r0, #0 moveq r0, ip, asr #31 orreq r0, r0, #1 RET 12: ARM_DIV2_ORDER r1, r2 cmp ip, #0 mov r0, r3, lsr r2 rsbmi r0, r0, #0 RET #endif /* ARM version */ DIV_FUNC_END divsi3 FUNC_START aeabi_idivmod #ifdef __thumb__ push {r0, r1, lr} bl SYM(__divsi3) POP {r1, r2, r3} mul r2, r0 sub r1, r1, r2 bx r3 #else stmfd sp!, { r0, r1, lr } bl SYM(__divsi3) ldmfd sp!, { r1, r2, lr } mul r3, r2, r0 sub r1, r1, r3 RET #endif FUNC_END aeabi_idivmod #endif /* L_divsi3 */ /* ------------------------------------------------------------------------ */ #ifdef L_modsi3 FUNC_START modsi3 #ifdef __thumb__ mov curbit, #1 cmp divisor, #0 beq LSYM(Ldiv0) bpl LSYM(Lover10) neg divisor, divisor @ Loops below use unsigned. LSYM(Lover10): push { work } @ Need to save the sign of the dividend, unfortunately, we need @ work later on. Must do this after saving the original value of @ the work register, because we will pop this value off first. push { dividend } cmp dividend, #0 bpl LSYM(Lover11) neg dividend, dividend LSYM(Lover11): cmp dividend, divisor blo LSYM(Lgot_result) THUMB_DIV_MOD_BODY 1 pop { work } cmp work, #0 bpl LSYM(Lover12) neg dividend, dividend LSYM(Lover12): pop { work } RET #else /* ARM version. */ cmp r1, #0 beq LSYM(Ldiv0) rsbmi r1, r1, #0 @ loops below use unsigned. movs ip, r0 @ preserve sign of dividend rsbmi r0, r0, #0 @ if negative make positive subs r2, r1, #1 @ compare divisor with 1 cmpne r0, r1 @ compare dividend with divisor moveq r0, #0 tsthi r1, r2 @ see if divisor is power of 2 andeq r0, r0, r2 bls 10f ARM_MOD_BODY r0, r1, r2, r3 10: cmp ip, #0 rsbmi r0, r0, #0 RET #endif /* ARM version */ DIV_FUNC_END modsi3 #endif /* L_modsi3 */ /* ------------------------------------------------------------------------ */ #ifdef L_dvmd_tls FUNC_START div0 FUNC_ALIAS aeabi_idiv0 div0 FUNC_ALIAS aeabi_ldiv0 div0 RET FUNC_END div0 #endif /* L_divmodsi_tools */ /* ------------------------------------------------------------------------ */ #ifdef L_dvmd_lnx @ GNU/Linux division-by zero handler. Used in place of L_dvmd_tls /* Constant taken from . */ #define SIGFPE 8 .code 32 FUNC_START div0 stmfd sp!, {r1, lr} mov r0, #SIGFPE bl SYM(raise) __PLT__ RETLDM r1 FUNC_END div0 #endif /* L_dvmd_lnx */ /* ------------------------------------------------------------------------ */ /* Dword shift operations. */ /* All the following Dword shift variants rely on the fact that shft xxx, Reg is in fact done as shft xxx, (Reg & 255) so for Reg value in (32...63) and (-1...-31) we will get zero (in the case of logical shifts) or the sign (for asr). */ #ifdef __ARMEB__ #define al r1 #define ah r0 #else #define al r0 #define ah r1 #endif /* Prevent __aeabi double-word shifts from being produced on SymbianOS. */ #ifndef __symbian__ #ifdef L_lshrdi3 FUNC_START lshrdi3 FUNC_ALIAS aeabi_llsr lshrdi3 #ifdef __thumb__ lsr al, r2 mov r3, ah lsr ah, r2 mov ip, r3 sub r2, #32 lsr r3, r2 orr al, r3 neg r2, r2 mov r3, ip lsl r3, r2 orr al, r3 RET #else subs r3, r2, #32 rsb ip, r2, #32 movmi al, al, lsr r2 movpl al, ah, lsr r3 orrmi al, al, ah, lsl ip mov ah, ah, lsr r2 RET #endif FUNC_END aeabi_llsr FUNC_END lshrdi3 #endif #ifdef L_ashrdi3 FUNC_START ashrdi3 FUNC_ALIAS aeabi_lasr ashrdi3 #ifdef __thumb__ lsr al, r2 mov r3, ah asr ah, r2 sub r2, #32 @ If r2 is negative at this point the following step would OR @ the sign bit into all of AL. That's not what we want... bmi 1f mov ip, r3 asr r3, r2 orr al, r3 mov r3, ip 1: neg r2, r2 lsl r3, r2 orr al, r3 RET #else subs r3, r2, #32 rsb ip, r2, #32 movmi al, al, lsr r2 movpl al, ah, asr r3 orrmi al, al, ah, lsl ip mov ah, ah, asr r2 RET #endif FUNC_END aeabi_lasr FUNC_END ashrdi3 #endif #ifdef L_ashldi3 FUNC_START ashldi3 FUNC_ALIAS aeabi_llsl ashldi3 #ifdef __thumb__ lsl ah, r2 mov r3, al lsl al, r2 mov ip, r3 sub r2, #32 lsl r3, r2 orr ah, r3 neg r2, r2 mov r3, ip lsr r3, r2 orr ah, r3 RET #else subs r3, r2, #32 rsb ip, r2, #32 movmi ah, ah, lsl r2 movpl ah, al, lsl r3 orrmi ah, ah, al, lsr ip mov al, al, lsl r2 RET #endif FUNC_END aeabi_llsl FUNC_END ashldi3 #endif #endif /* __symbian__ */ /* ------------------------------------------------------------------------ */ /* These next two sections are here despite the fact that they contain Thumb assembler because their presence allows interworked code to be linked even when the GCC library is this one. */ /* Do not build the interworking functions when the target architecture does not support Thumb instructions. (This can be a multilib option). */ #if defined __ARM_ARCH_4T__ || defined __ARM_ARCH_5T__\ || defined __ARM_ARCH_5TE__ || defined __ARM_ARCH_5TEJ__ \ || __ARM_ARCH__ >= 6 #if defined L_call_via_rX /* These labels & instructions are used by the Arm/Thumb interworking code. The address of function to be called is loaded into a register and then one of these labels is called via a BL instruction. This puts the return address into the link register with the bottom bit set, and the code here switches to the correct mode before executing the function. */ .text .align 0 .force_thumb .macro call_via register THUMB_FUNC_START _call_via_\register bx \register nop SIZE (_call_via_\register) .endm call_via r0 call_via r1 call_via r2 call_via r3 call_via r4 call_via r5 call_via r6 call_via r7 call_via r8 call_via r9 call_via sl call_via fp call_via ip call_via sp call_via lr #endif /* L_call_via_rX */ #if defined L_interwork_call_via_rX /* These labels & instructions are used by the Arm/Thumb interworking code, when the target address is in an unknown instruction set. The address of function to be called is loaded into a register and then one of these labels is called via a BL instruction. This puts the return address into the link register with the bottom bit set, and the code here switches to the correct mode before executing the function. Unfortunately the target code cannot be relied upon to return via a BX instruction, so instead we have to store the resturn address on the stack and allow the called function to return here instead. Upon return we recover the real return address and use a BX to get back to Thumb mode. There are three variations of this code. The first, _interwork_call_via_rN(), will push the return address onto the stack and pop it in _arm_return(). It should only be used if all arguments are passed in registers. The second, _interwork_r7_call_via_rN(), instead stores the return address at [r7, #-4]. It is the caller's responsibility to ensure that this address is valid and contains no useful data. The third, _interwork_r11_call_via_rN(), works in the same way but uses r11 instead of r7. It is useful if the caller does not really need a frame pointer. */ .text .align 0 .code 32 .globl _arm_return LSYM(Lstart_arm_return): cfi_start LSYM(Lstart_arm_return) LSYM(Lend_arm_return) cfi_push 0, 0xe, -0x8, 0x8 nop @ This nop is for the benefit of debuggers, so that @ backtraces will use the correct unwind information. _arm_return: RETLDM unwind=LSYM(Lstart_arm_return) cfi_end LSYM(Lend_arm_return) .globl _arm_return_r7 _arm_return_r7: ldr lr, [r7, #-4] bx lr .globl _arm_return_r11 _arm_return_r11: ldr lr, [r11, #-4] bx lr .macro interwork_with_frame frame, register, name, return .code 16 THUMB_FUNC_START \name bx pc nop .code 32 tst \register, #1 streq lr, [\frame, #-4] adreq lr, _arm_return_\frame bx \register SIZE (\name) .endm .macro interwork register .code 16 THUMB_FUNC_START _interwork_call_via_\register bx pc nop .code 32 .globl LSYM(Lchange_\register) LSYM(Lchange_\register): tst \register, #1 streq lr, [sp, #-8]! adreq lr, _arm_return bx \register SIZE (_interwork_call_via_\register) interwork_with_frame r7,\register,_interwork_r7_call_via_\register interwork_with_frame r11,\register,_interwork_r11_call_via_\register .endm interwork r0 interwork r1 interwork r2 interwork r3 interwork r4 interwork r5 interwork r6 interwork r7 interwork r8 interwork r9 interwork sl interwork fp interwork ip interwork sp /* The LR case has to be handled a little differently... */ .code 16 THUMB_FUNC_START _interwork_call_via_lr bx pc nop .code 32 .globl .Lchange_lr .Lchange_lr: tst lr, #1 stmeqdb r13!, {lr, pc} mov ip, lr adreq lr, _arm_return bx ip SIZE (_interwork_call_via_lr) #endif /* L_interwork_call_via_rX */ #endif /* Arch supports thumb. */ #ifndef __symbian__ #include "ieee754-df.S" #include "ieee754-sf.S" #include "bpabi.S" #endif /* __symbian__ */ Index: projects/clang-trunk/contrib/gcc =================================================================== --- projects/clang-trunk/contrib/gcc (revision 283870) +++ projects/clang-trunk/contrib/gcc (revision 283871) Property changes on: projects/clang-trunk/contrib/gcc ___________________________________________________________________ Modified: svn:mergeinfo ## -0,0 +0,1 ## Merged /head/contrib/gcc:r283596-283870 Index: projects/clang-trunk/contrib/netbsd-tests/lib/libc/c063/t_faccessat.c =================================================================== --- projects/clang-trunk/contrib/netbsd-tests/lib/libc/c063/t_faccessat.c (revision 283870) +++ projects/clang-trunk/contrib/netbsd-tests/lib/libc/c063/t_faccessat.c (revision 283871) @@ -1,188 +1,191 @@ /* $NetBSD: t_faccessat.c,v 1.2 2013/03/17 04:46:06 jmmv Exp $ */ /*- * Copyright (c) 2012 The NetBSD Foundation, Inc. * All rights reserved. * * This code is derived from software contributed to The NetBSD Foundation * by Emmanuel Dreyfus. * * Redistribution and use in source and binary forms, with or without * modification, are permitted provided that the following conditions * are met: * 1. Redistributions of source code must retain the above copyright * notice, this list of conditions and the following disclaimer. * 2. Redistributions in binary form must reproduce the above copyright * notice, this list of conditions and the following disclaimer in the * documentation and/or other materials provided with the distribution. * * THIS SOFTWARE IS PROVIDED BY THE NETBSD FOUNDATION, INC. AND CONTRIBUTORS * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR * PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE FOUNDATION OR CONTRIBUTORS * BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE * POSSIBILITY OF SUCH DAMAGE. */ #include __RCSID("$NetBSD: t_faccessat.c,v 1.2 2013/03/17 04:46:06 jmmv Exp $"); #include #include #include #include #include #include #include #include #include +#ifdef __FreeBSD__ +#include +#endif #define DIR "dir" #define FILE "dir/faccessat" #define BASEFILE "faccessat" #define LINK "dir/symlink" #define BASELINK "symlink" #define FILEERR "dir/faccessaterr" ATF_TC(faccessat_fd); ATF_TC_HEAD(faccessat_fd, tc) { atf_tc_set_md_var(tc, "descr", "See that faccessat works with fd"); } ATF_TC_BODY(faccessat_fd, tc) { int dfd; int fd; ATF_REQUIRE(mkdir(DIR, 0755) == 0); ATF_REQUIRE((fd = open(FILE, O_CREAT|O_RDWR, 0644)) != -1); ATF_REQUIRE(close(fd) == 0); ATF_REQUIRE((dfd = open(DIR, O_RDONLY, 0)) != -1); ATF_REQUIRE(faccessat(dfd, BASEFILE, F_OK, 0) == 0); ATF_REQUIRE(close(dfd) == 0); } ATF_TC(faccessat_fdcwd); ATF_TC_HEAD(faccessat_fdcwd, tc) { atf_tc_set_md_var(tc, "descr", "See that faccessat works with fd as AT_FDCWD"); } ATF_TC_BODY(faccessat_fdcwd, tc) { int fd; ATF_REQUIRE(mkdir(DIR, 0755) == 0); ATF_REQUIRE((fd = open(FILE, O_CREAT|O_RDWR, 0644)) != -1); ATF_REQUIRE(close(fd) == 0); ATF_REQUIRE(chdir(DIR) == 0); ATF_REQUIRE(faccessat(AT_FDCWD, BASEFILE, F_OK, 0) == 0); } ATF_TC(faccessat_fdcwderr); ATF_TC_HEAD(faccessat_fdcwderr, tc) { atf_tc_set_md_var(tc, "descr", "See that faccessat fails with fd as AT_FDCWD and bad path"); } ATF_TC_BODY(faccessat_fdcwderr, tc) { ATF_REQUIRE(mkdir(DIR, 0755) == 0); ATF_REQUIRE(faccessat(AT_FDCWD, FILEERR, F_OK, 0) == -1); } ATF_TC(faccessat_fderr1); ATF_TC_HEAD(faccessat_fderr1, tc) { atf_tc_set_md_var(tc, "descr", "See that faccessat fail with bad path"); } ATF_TC_BODY(faccessat_fderr1, tc) { int dfd; ATF_REQUIRE(mkdir(DIR, 0755) == 0); ATF_REQUIRE((dfd = open(DIR, O_RDONLY, 0)) != -1); ATF_REQUIRE(faccessat(dfd, FILEERR, F_OK, 0) == -1); ATF_REQUIRE(close(dfd) == 0); } ATF_TC(faccessat_fderr2); ATF_TC_HEAD(faccessat_fderr2, tc) { atf_tc_set_md_var(tc, "descr", "See that faccessat fails with bad fdat"); } ATF_TC_BODY(faccessat_fderr2, tc) { int dfd; int fd; char cwd[MAXPATHLEN]; ATF_REQUIRE(mkdir(DIR, 0755) == 0); ATF_REQUIRE((fd = open(FILE, O_CREAT|O_RDWR, 0644)) != -1); ATF_REQUIRE(close(fd) == 0); ATF_REQUIRE((dfd = open(getcwd(cwd, MAXPATHLEN), O_RDONLY, 0)) != -1); ATF_REQUIRE(faccessat(dfd, BASEFILE, F_OK, 0) == -1); ATF_REQUIRE(close(dfd) == 0); } ATF_TC(faccessat_fderr3); ATF_TC_HEAD(faccessat_fderr3, tc) { atf_tc_set_md_var(tc, "descr", "See that faccessat fails with fd as -1"); } ATF_TC_BODY(faccessat_fderr3, tc) { int fd; ATF_REQUIRE(mkdir(DIR, 0755) == 0); ATF_REQUIRE((fd = open(FILE, O_CREAT|O_RDWR, 0644)) != -1); ATF_REQUIRE(close(fd) == 0); ATF_REQUIRE(faccessat(-1, FILE, F_OK, 0) == -1); } ATF_TC(faccessat_fdlink); ATF_TC_HEAD(faccessat_fdlink, tc) { atf_tc_set_md_var(tc, "descr", "See that faccessat works on symlink"); } ATF_TC_BODY(faccessat_fdlink, tc) { int dfd; ATF_REQUIRE(mkdir(DIR, 0755) == 0); ATF_REQUIRE(symlink(FILE, LINK) == 0); /* NB: FILE does not exists */ ATF_REQUIRE((dfd = open(DIR, O_RDONLY, 0)) != -1); ATF_REQUIRE(faccessat(dfd, BASELINK, F_OK, 0) == -1); ATF_REQUIRE(errno == ENOENT); #ifdef __FreeBSD__ atf_tc_expect_fail("Depends on non-standard behavior not mentioned in POSIX.1-2008"); #endif ATF_REQUIRE(faccessat(dfd, BASELINK, F_OK, AT_SYMLINK_NOFOLLOW) == 0); ATF_REQUIRE(close(dfd) == 0); } ATF_TP_ADD_TCS(tp) { ATF_TP_ADD_TC(tp, faccessat_fd); ATF_TP_ADD_TC(tp, faccessat_fdcwd); ATF_TP_ADD_TC(tp, faccessat_fdcwderr); ATF_TP_ADD_TC(tp, faccessat_fderr1); ATF_TP_ADD_TC(tp, faccessat_fderr2); ATF_TP_ADD_TC(tp, faccessat_fderr3); ATF_TP_ADD_TC(tp, faccessat_fdlink); return atf_no_error(); } Index: projects/clang-trunk/contrib/netbsd-tests/lib/libc/c063/t_openat.c =================================================================== --- projects/clang-trunk/contrib/netbsd-tests/lib/libc/c063/t_openat.c (revision 283870) +++ projects/clang-trunk/contrib/netbsd-tests/lib/libc/c063/t_openat.c (revision 283871) @@ -1,165 +1,168 @@ /* $NetBSD: t_openat.c,v 1.2 2013/03/17 04:46:06 jmmv Exp $ */ /*- * Copyright (c) 2012 The NetBSD Foundation, Inc. * All rights reserved. * * This code is derived from software contributed to The NetBSD Foundation * by Emmanuel Dreyfus. * * Redistribution and use in source and binary forms, with or without * modification, are permitted provided that the following conditions * are met: * 1. Redistributions of source code must retain the above copyright * notice, this list of conditions and the following disclaimer. * 2. Redistributions in binary form must reproduce the above copyright * notice, this list of conditions and the following disclaimer in the * documentation and/or other materials provided with the distribution. * * THIS SOFTWARE IS PROVIDED BY THE NETBSD FOUNDATION, INC. AND CONTRIBUTORS * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR * PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE FOUNDATION OR CONTRIBUTORS * BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE * POSSIBILITY OF SUCH DAMAGE. */ #include __RCSID("$NetBSD: t_openat.c,v 1.2 2013/03/17 04:46:06 jmmv Exp $"); #include #include #include #include #include #include #include #include #include +#ifdef __FreeBSD__ +#include +#endif #define DIR "dir" #define FILE "dir/openat" #define BASEFILE "openat" #define FILEERR "dir/openaterr" ATF_TC(openat_fd); ATF_TC_HEAD(openat_fd, tc) { atf_tc_set_md_var(tc, "descr", "See that openat works with fd"); } ATF_TC_BODY(openat_fd, tc) { int dfd; int fd; ATF_REQUIRE(mkdir(DIR, 0755) == 0); ATF_REQUIRE((fd = open(FILE, O_CREAT|O_RDWR, 0644)) != -1); ATF_REQUIRE(close(fd) == 0); ATF_REQUIRE((dfd = open(DIR, O_RDONLY, 0)) != -1); ATF_REQUIRE((fd = openat(dfd, BASEFILE, O_RDONLY, 0)) != -1); ATF_REQUIRE(close(dfd) == 0); ATF_REQUIRE(close(fd) == 0); } ATF_TC(openat_fdcwd); ATF_TC_HEAD(openat_fdcwd, tc) { atf_tc_set_md_var(tc, "descr", "See that openat works with fd as AT_FDCWD"); } ATF_TC_BODY(openat_fdcwd, tc) { int fd; ATF_REQUIRE(mkdir(DIR, 0755) == 0); ATF_REQUIRE((fd = open(FILE, O_CREAT|O_RDWR, 0644)) != -1); ATF_REQUIRE(close(fd) == 0); ATF_REQUIRE(chdir(DIR) == 0); ATF_REQUIRE((fd = openat(AT_FDCWD, BASEFILE, O_RDONLY, 0)) != -1); ATF_REQUIRE(close(fd) == 0); } ATF_TC(openat_fdcwderr); ATF_TC_HEAD(openat_fdcwderr, tc) { atf_tc_set_md_var(tc, "descr", "See that openat fails with fd as AT_FDCWD and bad path"); } ATF_TC_BODY(openat_fdcwderr, tc) { int fd; ATF_REQUIRE(mkdir(DIR, 0755) == 0); ATF_REQUIRE((fd = openat(AT_FDCWD, FILEERR, O_RDONLY, 0)) == -1); } ATF_TC(openat_fderr1); ATF_TC_HEAD(openat_fderr1, tc) { atf_tc_set_md_var(tc, "descr", "See that openat fail with bad path"); } ATF_TC_BODY(openat_fderr1, tc) { int dfd; int fd; ATF_REQUIRE(mkdir(DIR, 0755) == 0); ATF_REQUIRE((dfd = open(DIR, O_RDONLY, 0)) != -1); ATF_REQUIRE((fd = openat(dfd, FILEERR, O_RDONLY, 0)) == -1); ATF_REQUIRE(close(dfd) == 0); } ATF_TC(openat_fderr2); ATF_TC_HEAD(openat_fderr2, tc) { atf_tc_set_md_var(tc, "descr", "See that openat fails with bad fdat"); } ATF_TC_BODY(openat_fderr2, tc) { int dfd; int fd; char cwd[MAXPATHLEN]; ATF_REQUIRE(mkdir(DIR, 0755) == 0); ATF_REQUIRE((fd = open(FILE, O_CREAT|O_RDWR, 0644)) != -1); ATF_REQUIRE(close(fd) == 0); ATF_REQUIRE((dfd = open(getcwd(cwd, MAXPATHLEN), O_RDONLY, 0)) != -1); ATF_REQUIRE((fd = openat(dfd, BASEFILE, O_RDONLY, 0)) == -1); ATF_REQUIRE(close(dfd) == 0); } ATF_TC(openat_fderr3); ATF_TC_HEAD(openat_fderr3, tc) { atf_tc_set_md_var(tc, "descr", "See that openat fails with fd as -1"); } ATF_TC_BODY(openat_fderr3, tc) { int fd; ATF_REQUIRE(mkdir(DIR, 0755) == 0); ATF_REQUIRE((fd = open(FILE, O_CREAT|O_RDWR, 0644)) != -1); ATF_REQUIRE(close(fd) == 0); ATF_REQUIRE((fd = openat(-1, FILE, O_RDONLY, 0)) == -1); } ATF_TP_ADD_TCS(tp) { ATF_TP_ADD_TC(tp, openat_fd); ATF_TP_ADD_TC(tp, openat_fdcwd); ATF_TP_ADD_TC(tp, openat_fdcwderr); ATF_TP_ADD_TC(tp, openat_fderr1); ATF_TP_ADD_TC(tp, openat_fderr2); ATF_TP_ADD_TC(tp, openat_fderr3); return atf_no_error(); } Index: projects/clang-trunk/contrib/netbsd-tests/lib/libc/c063/t_readlinkat.c =================================================================== --- projects/clang-trunk/contrib/netbsd-tests/lib/libc/c063/t_readlinkat.c (revision 283870) +++ projects/clang-trunk/contrib/netbsd-tests/lib/libc/c063/t_readlinkat.c (revision 283871) @@ -1,157 +1,160 @@ /* $NetBSD: t_readlinkat.c,v 1.3 2013/03/17 04:46:06 jmmv Exp $ */ /*- * Copyright (c) 2012 The NetBSD Foundation, Inc. * All rights reserved. * * This code is derived from software contributed to The NetBSD Foundation * by Emmanuel Dreyfus. * * Redistribution and use in source and binary forms, with or without * modification, are permitted provided that the following conditions * are met: * 1. Redistributions of source code must retain the above copyright * notice, this list of conditions and the following disclaimer. * 2. Redistributions in binary form must reproduce the above copyright * notice, this list of conditions and the following disclaimer in the * documentation and/or other materials provided with the distribution. * * THIS SOFTWARE IS PROVIDED BY THE NETBSD FOUNDATION, INC. AND CONTRIBUTORS * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR * PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE FOUNDATION OR CONTRIBUTORS * BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE * POSSIBILITY OF SUCH DAMAGE. */ #include __RCSID("$NetBSD: t_readlinkat.c,v 1.3 2013/03/17 04:46:06 jmmv Exp $"); #include #include #include #include #include #include #include #include #include +#ifdef __FreeBSD__ +#include +#endif #define DIR "dir" #define FILE "dir/readlinkat" #define BASEFILE "readlinkat" #define LINK "dir/symlink" #define BASELINK "symlink" #define FILEERR "dir/readlinkaterr" ATF_TC(readlinkat_fd); ATF_TC_HEAD(readlinkat_fd, tc) { atf_tc_set_md_var(tc, "descr", "See that readlinkat works with fd"); } ATF_TC_BODY(readlinkat_fd, tc) { int dfd; int fd; ssize_t len; char buf[MAXPATHLEN]; ATF_REQUIRE(mkdir(DIR, 0755) == 0); ATF_REQUIRE((fd = open(FILE, O_CREAT|O_RDWR, 0644)) != -1); ATF_REQUIRE(close(fd) == 0); ATF_REQUIRE(symlink(FILE, LINK) == 0); ATF_REQUIRE((dfd = open(DIR, O_RDONLY, 0)) != -1); len = readlinkat(dfd, BASELINK, buf, sizeof(buf)-1); ATF_REQUIRE(len != -1); buf[len] = 0; ATF_REQUIRE(close(dfd) == 0); ATF_REQUIRE(strcmp(buf, FILE) == 0); } ATF_TC(readlinkat_fdcwd); ATF_TC_HEAD(readlinkat_fdcwd, tc) { atf_tc_set_md_var(tc, "descr", "See that readlinkat works with fd as AT_FDCWD"); } ATF_TC_BODY(readlinkat_fdcwd, tc) { int fd; ssize_t len; char buf[MAXPATHLEN]; ATF_REQUIRE(mkdir(DIR, 0755) == 0); ATF_REQUIRE((fd = open(FILE, O_CREAT|O_RDWR, 0644)) != -1); ATF_REQUIRE(close(fd) == 0); ATF_REQUIRE(symlink(FILE, LINK) == 0); len = readlinkat(AT_FDCWD, LINK, buf, sizeof(buf)-1); ATF_REQUIRE(len != -1); buf[len] = 0; ATF_REQUIRE(strcmp(buf, FILE) == 0); } ATF_TC(readlinkat_fdcwderr); ATF_TC_HEAD(readlinkat_fdcwderr, tc) { atf_tc_set_md_var(tc, "descr", "See that readlinkat fails with fd as AT_FDCWD and bad path"); } ATF_TC_BODY(readlinkat_fdcwderr, tc) { char buf[MAXPATHLEN]; ATF_REQUIRE(readlinkat(AT_FDCWD, LINK, buf, sizeof(buf)) == -1); } ATF_TC(readlinkat_fderr1); ATF_TC_HEAD(readlinkat_fderr1, tc) { atf_tc_set_md_var(tc, "descr", "See that readlinkat fail with bad path"); } ATF_TC_BODY(readlinkat_fderr1, tc) { int dfd; ATF_REQUIRE(mkdir(DIR, 0755) == 0); ATF_REQUIRE((dfd = open(DIR, O_RDONLY, 0)) != -1); ATF_REQUIRE(readlinkat(dfd, FILEERR, F_OK, 0) == -1); ATF_REQUIRE(close(dfd) == 0); } ATF_TC(readlinkat_fderr2); ATF_TC_HEAD(readlinkat_fderr2, tc) { atf_tc_set_md_var(tc, "descr", "See that readlinkat fails with fd as -1"); } ATF_TC_BODY(readlinkat_fderr2, tc) { int fd; char buf[MAXPATHLEN]; ATF_REQUIRE(mkdir(DIR, 0755) == 0); ATF_REQUIRE((fd = open(FILE, O_CREAT|O_RDWR, 0644)) != -1); ATF_REQUIRE(close(fd) == 0); ATF_REQUIRE(symlink(FILE, LINK) == 0); ATF_REQUIRE(readlinkat(-1, LINK, buf, sizeof(buf)) == -1); } ATF_TP_ADD_TCS(tp) { ATF_TP_ADD_TC(tp, readlinkat_fd); ATF_TP_ADD_TC(tp, readlinkat_fdcwd); ATF_TP_ADD_TC(tp, readlinkat_fdcwderr); ATF_TP_ADD_TC(tp, readlinkat_fderr1); ATF_TP_ADD_TC(tp, readlinkat_fderr2); return atf_no_error(); } Index: projects/clang-trunk/contrib/netbsd-tests/lib/libc/c063/t_unlinkat.c =================================================================== --- projects/clang-trunk/contrib/netbsd-tests/lib/libc/c063/t_unlinkat.c (revision 283870) +++ projects/clang-trunk/contrib/netbsd-tests/lib/libc/c063/t_unlinkat.c (revision 283871) @@ -1,176 +1,179 @@ /* $NetBSD: t_unlinkat.c,v 1.2 2013/03/17 04:46:06 jmmv Exp $ */ /*- * Copyright (c) 2012 The NetBSD Foundation, Inc. * All rights reserved. * * This code is derived from software contributed to The NetBSD Foundation * by Emmanuel Dreyfus. * * Redistribution and use in source and binary forms, with or without * modification, are permitted provided that the following conditions * are met: * 1. Redistributions of source code must retain the above copyright * notice, this list of conditions and the following disclaimer. * 2. Redistributions in binary form must reproduce the above copyright * notice, this list of conditions and the following disclaimer in the * documentation and/or other materials provided with the distribution. * * THIS SOFTWARE IS PROVIDED BY THE NETBSD FOUNDATION, INC. AND CONTRIBUTORS * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR * PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE FOUNDATION OR CONTRIBUTORS * BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE * POSSIBILITY OF SUCH DAMAGE. */ #include __RCSID("$NetBSD: t_unlinkat.c,v 1.2 2013/03/17 04:46:06 jmmv Exp $"); #include #include #include #include #include #include #include #include #include +#ifdef __FreeBSD__ +#include +#endif #define DIR "dir" #define FILE "dir/unlinkat" #define BASEFILE "unlinkat" #define FILEERR "dir/unlinkaterr" ATF_TC(unlinkat_fd); ATF_TC_HEAD(unlinkat_fd, tc) { atf_tc_set_md_var(tc, "descr", "See that unlinkat works with fd"); } ATF_TC_BODY(unlinkat_fd, tc) { int dfd; int fd; ATF_REQUIRE(mkdir(DIR, 0755) == 0); ATF_REQUIRE((fd = open(FILE, O_CREAT|O_RDWR, 0644)) != -1); ATF_REQUIRE(close(fd) == 0); ATF_REQUIRE((dfd = open(DIR, O_RDONLY, 0)) != -1); ATF_REQUIRE(unlinkat(dfd, BASEFILE, 0) == 0); ATF_REQUIRE(close(dfd) == 0); } ATF_TC(unlinkat_fdcwd); ATF_TC_HEAD(unlinkat_fdcwd, tc) { atf_tc_set_md_var(tc, "descr", "See that unlinkat works with fd as AT_FDCWD"); } ATF_TC_BODY(unlinkat_fdcwd, tc) { int fd; ATF_REQUIRE(mkdir(DIR, 0755) == 0); ATF_REQUIRE((fd = open(FILE, O_CREAT|O_RDWR, 0644)) != -1); ATF_REQUIRE(close(fd) == 0); ATF_REQUIRE(chdir(DIR) == 0); ATF_REQUIRE(unlinkat(AT_FDCWD, BASEFILE, 0) == 0); } ATF_TC(unlinkat_fdcwderr); ATF_TC_HEAD(unlinkat_fdcwderr, tc) { atf_tc_set_md_var(tc, "descr", "See that unlinkat fails with fd as AT_FDCWD and bad path"); } ATF_TC_BODY(unlinkat_fdcwderr, tc) { ATF_REQUIRE(mkdir(DIR, 0755) == 0); ATF_REQUIRE(unlinkat(AT_FDCWD, FILEERR, 0) == -1); } ATF_TC(unlinkat_fderr1); ATF_TC_HEAD(unlinkat_fderr1, tc) { atf_tc_set_md_var(tc, "descr", "See that unlinkat fail with bad path"); } ATF_TC_BODY(unlinkat_fderr1, tc) { int dfd; ATF_REQUIRE(mkdir(DIR, 0755) == 0); ATF_REQUIRE((dfd = open(DIR, O_RDONLY, 0)) != -1); ATF_REQUIRE(unlinkat(dfd, FILEERR, 0) == -1); ATF_REQUIRE(close(dfd) == 0); } ATF_TC(unlinkat_fderr2); ATF_TC_HEAD(unlinkat_fderr2, tc) { atf_tc_set_md_var(tc, "descr", "See that unlinkat fails with bad fdat"); } ATF_TC_BODY(unlinkat_fderr2, tc) { int dfd; int fd; char cwd[MAXPATHLEN]; ATF_REQUIRE(mkdir(DIR, 0755) == 0); ATF_REQUIRE((fd = open(FILE, O_CREAT|O_RDWR, 0644)) != -1); ATF_REQUIRE(close(fd) == 0); ATF_REQUIRE((dfd = open(getcwd(cwd, MAXPATHLEN), O_RDONLY, 0)) != -1); ATF_REQUIRE(unlinkat(dfd, BASEFILE, 0) == -1); ATF_REQUIRE(close(dfd) == 0); } ATF_TC(unlinkat_fderr3); ATF_TC_HEAD(unlinkat_fderr3, tc) { atf_tc_set_md_var(tc, "descr", "See that unlinkat fails with fd as -1"); } ATF_TC_BODY(unlinkat_fderr3, tc) { int fd; ATF_REQUIRE(mkdir(DIR, 0755) == 0); ATF_REQUIRE((fd = open(FILE, O_CREAT|O_RDWR, 0644)) != -1); ATF_REQUIRE(close(fd) == 0); ATF_REQUIRE(unlinkat(-1, FILE, 0) == -1); } ATF_TC(unlinkat_dir); ATF_TC_HEAD(unlinkat_dir, tc) { atf_tc_set_md_var(tc, "descr", "See that unlinkat can remove directories"); } ATF_TC_BODY(unlinkat_dir, tc) { ATF_REQUIRE(mkdir(DIR, 0755) == 0); ATF_REQUIRE(unlinkat(AT_FDCWD, DIR, 0) == -1); ATF_REQUIRE(errno == EPERM); ATF_REQUIRE(unlinkat(AT_FDCWD, DIR, AT_REMOVEDIR) == 0); } ATF_TP_ADD_TCS(tp) { ATF_TP_ADD_TC(tp, unlinkat_fd); ATF_TP_ADD_TC(tp, unlinkat_fdcwd); ATF_TP_ADD_TC(tp, unlinkat_fdcwderr); ATF_TP_ADD_TC(tp, unlinkat_fderr1); ATF_TP_ADD_TC(tp, unlinkat_fderr2); ATF_TP_ADD_TC(tp, unlinkat_fderr3); ATF_TP_ADD_TC(tp, unlinkat_dir); return atf_no_error(); } Index: projects/clang-trunk/contrib/traceroute/traceroute.8 =================================================================== --- projects/clang-trunk/contrib/traceroute/traceroute.8 (revision 283870) +++ projects/clang-trunk/contrib/traceroute/traceroute.8 (revision 283871) @@ -1,395 +1,398 @@ .\" Copyright (c) 1989, 1995, 1996, 1997, 1999, 2000 .\" The Regents of the University of California. All rights reserved. .\" .\" Redistribution and use in source and binary forms are permitted .\" provided that the above copyright notice and this paragraph are .\" duplicated in all such forms and that any documentation, .\" advertising materials, and other materials related to such .\" distribution and use acknowledge that the software was developed .\" by the University of California, Berkeley. The name of the .\" University may not be used to endorse or promote products derived .\" from this software without specific prior written permission. .\" THIS SOFTWARE IS PROVIDED ``AS IS'' AND WITHOUT ANY EXPRESS OR .\" IMPLIED WARRANTIES, INCLUDING, WITHOUT LIMITATION, THE IMPLIED .\" WARRANTIES OF MERCHANTIBILITY AND FITNESS FOR A PARTICULAR PURPOSE. .\" .\" $Id: traceroute.8,v 1.19 2000/09/21 08:44:19 leres Exp $ .\" $FreeBSD$ .\" -.Dd June 19, 2012 +.Dd May 31, 2015 .Dt TRACEROUTE 8 .Os .Sh NAME .Nm traceroute .Nd "print the route packets take to network host" .Sh SYNOPSIS .Nm .Bk -words .Op Fl adDeFISnrvx .Op Fl f Ar first_ttl .Op Fl g Ar gateway .Op Fl M Ar first_ttl .Op Fl m Ar max_ttl .Op Fl P Ar proto .Op Fl p Ar port .Op Fl q Ar nqueries .Op Fl s Ar src_addr .Op Fl t Ar tos .Op Fl w Ar waittime .Op Fl A Ar as_server .Op Fl z Ar pausemsecs .Ar host .Op Ar packetlen .Ek .Sh DESCRIPTION The Internet is a large and complex aggregation of network hardware, connected together by gateways. Tracking the route one's packets follow (or finding the miscreant gateway that's discarding your packets) can be difficult. .Nm utilizes the IP protocol `time to live' field and attempts to elicit an ICMP TIME_EXCEEDED response from each gateway along the path to some host. .Pp The only mandatory parameter is the destination host name or IP number. The default probe datagram length is 40 bytes, but this may be increased by specifying a packet length (in bytes) after the destination host name. .Pp Other options are: .Bl -tag -width Ds .It Fl a Turn on AS# lookups for each hop encountered. .It Fl A Ar as_server Turn on AS# lookups and use the given server instead of the default. .It Fl e Firewall evasion mode. -Use fixed destination ports for UDP and TCP probes. +Use fixed destination ports for UDP, UDP-Lite, TCP and SCTP probes. The destination port does NOT increment with each packet sent. .It Fl f Ar first_ttl Set the initial time-to-live used in the first outgoing probe packet. .It Fl F Set the "don't fragment" bit. .It Fl d Enable socket level debugging. .It Fl D When an ICMP response to our probe datagram is received, print the differences between the transmitted packet and the packet quoted by the ICMP response. A key showing the location of fields within the transmitted packet is printed, followed by the original packet in hex, followed by the quoted packet in hex. Bytes that are unchanged in the quoted packet are shown as underscores. Note, the IP checksum and the TTL of the quoted packet are not expected to match. By default, only one probe per hop is sent with this option. .It Fl g Ar gateway Specify a loose source route gateway (8 maximum). .It Fl i Ar iface Specify a network interface to obtain the source IP address for outgoing probe packets. This is normally only useful on a multi-homed host. (See the .Fl s flag for another way to do this.) .It Fl I Use ICMP ECHO instead of UDP datagrams. (A synonym for "-P icmp"). .It Fl M Ar first_ttl Set the initial time-to-live value used in outgoing probe packets. The default is 1, i.e., start with the first hop. .It Fl m Ar max_ttl Set the max time-to-live (max number of hops) used in outgoing probe packets. The default is the value of the .Va net.inet.ip.ttl .Xr sysctl 8 (the same default used for TCP connections). .It Fl n Print hop addresses numerically rather than symbolically and numerically (saves a nameserver address-to-name lookup for each gateway found on the path). .It Fl P Ar proto Send packets of specified IP protocol. The currently supported protocols -are: UDP, TCP, GRE and ICMP. Other protocols may also be specified (either by -name or by number), though +are: UDP, UDP-Lite, TCP, SCTP, GRE and ICMP. Other protocols may also be +specified (either by name or by number), though .Nm does not implement any special knowledge of their packet formats. This option is useful for determining which router along a path may be blocking packets based on IP protocol number. But see BUGS below. .It Fl p Ar port -Protocol specific. For UDP and TCP, sets +Protocol specific. For UDP, UDP-Lite, TCP and SCTP, sets the base .Ar port number used in probes (default is 33434). -Traceroute hopes that nothing is listening on UDP ports -.Em base +Traceroute hopes that nothing is listening on UDP ports (or UDP-Lite ports +if used by +.Nm +and supported by the peer) +.Em base + 1 to -.Em base + nhops * nprobes - 1 +.Em base + nhops * nprobes at the destination host (so an ICMP PORT_UNREACHABLE message will be returned to terminate the route tracing). If something is listening on a port in the default range, this option can be used to pick an unused port range. .It Fl q Ar nqueries Set the number of probes per hop (default is 3, unless .Fl D is specified, when it is 1). .It Fl r Bypass the normal routing tables and send directly to a host on an attached network. If the host is not on a directly-attached network, an error is returned. This option can be used to ping a local host through an interface that has no route through it (e.g., after the interface was dropped by .Xr routed 8 . .It Fl s Ar src_addr Use the following IP address (which usually is given as an IP number, not a hostname) as the source address in outgoing probe packets. On multi-homed hosts (those with more than one IP address), this option can be used to force the source address to be something other than the IP address of the interface the probe packet is sent on. If the IP address is not one of this machine's interface addresses, an error is returned and nothing is sent. (See the .Fl i flag for another way to do this.) .It Fl S Print a summary of how many probes were not answered for each hop. .It Fl t Ar tos Set the .Em type-of-service in probe packets to the following value (default zero). The value must be a decimal integer in the range 0 to 255. This option can be used to see if different types-of-service result in different paths. (If you are not running 4.4bsd, this may be academic since the normal network services like telnet and ftp don't let you control the TOS). Not all values of TOS are legal or meaningful \- see the IP spec for definitions. Useful values are probably .Fl t Ar 16 (low delay) and .Fl t Ar 8 (high throughput). .It Fl v Verbose output. Received ICMP packets other than .Dv TIME_EXCEEDED and .Dv UNREACHABLE Ns s are listed. .It Fl w Ar waittime Set the time (in seconds) to wait for a response to a probe (default 5 sec.). .It Fl x Toggle ip checksums. Normally, this prevents traceroute from calculating ip checksums. In some cases, the operating system can overwrite parts of the outgoing packet but not recalculate the checksum (so in some cases the default is to not calculate checksums and using .Fl x causes them to be calculated). Note that checksums are usually required for the last hop when using ICMP ECHO probes .Pq Fl I . So they are always calculated when using ICMP. .It Fl z Ar pausemsecs Set the time (in milliseconds) to pause between probes (default 0). Some systems such as Solaris and routers such as Ciscos rate limit icmp messages. A good value to use with this this is 500 (e.g. 1/2 second). .El .Pp This program attempts to trace the route an IP packet would follow to some internet host by launching UDP probe packets with a small ttl (time to live) then listening for an ICMP "time exceeded" reply from a gateway. We start our probes with a ttl of one and increase by one until we get an ICMP "port unreachable" (which means we got to "host") or hit a max (which defaults to the amount of hops specified by the .Va net.inet.ip.ttl .Xr sysctl 8 and can be changed with the .Fl m flag). Three probes (change with .Fl q flag) are sent at each ttl setting and a line is printed showing the ttl, address of the gateway and round trip time of each probe. If the probe answers come from different gateways, the address of each responding system will be printed. If there is no response within a 5 sec. timeout interval (changed with the .Fl w flag), a "*" is printed for that probe. .Pp We don't want the destination host to process the UDP probe packets so the destination port is set to an unlikely value (if some clod on the destination is using that value, it can be changed with the .Fl p flag). .Pp A sample use and output might be: .Bd -literal -offset 4n % traceroute nis.nsf.net. traceroute to nis.nsf.net (35.1.1.48), 64 hops max, 38 byte packet 1 helios.ee.lbl.gov (128.3.112.1) 19 ms 19 ms 0 ms 2 lilac-dmc.Berkeley.EDU (128.32.216.1) 39 ms 39 ms 19 ms 3 lilac-dmc.Berkeley.EDU (128.32.216.1) 39 ms 39 ms 19 ms 4 ccngw-ner-cc.Berkeley.EDU (128.32.136.23) 39 ms 40 ms 39 ms 5 ccn-nerif22.Berkeley.EDU (128.32.168.22) 39 ms 39 ms 39 ms 6 128.32.197.4 (128.32.197.4) 40 ms 59 ms 59 ms 7 131.119.2.5 (131.119.2.5) 59 ms 59 ms 59 ms 8 129.140.70.13 (129.140.70.13) 99 ms 99 ms 80 ms 9 129.140.71.6 (129.140.71.6) 139 ms 239 ms 319 ms 10 129.140.81.7 (129.140.81.7) 220 ms 199 ms 199 ms 11 nic.merit.edu (35.1.1.48) 239 ms 239 ms 239 ms .Ed .Pp Note that lines 2 & 3 are the same. This is due to a buggy kernel on the 2nd hop system \- lilac-dmc.Berkeley.EDU \- that forwards packets with a zero ttl (a bug in the distributed version of 4.3BSD). Note that you have to guess what path the packets are taking cross-country since the NSFNet (129.140) doesn't supply address-to-name translations for its NSSes. .Pp A more interesting example is: .Bd -literal -offset 4n % traceroute allspice.lcs.mit.edu. traceroute to allspice.lcs.mit.edu (18.26.0.115), 64 hops max 1 helios.ee.lbl.gov (128.3.112.1) 0 ms 0 ms 0 ms 2 lilac-dmc.Berkeley.EDU (128.32.216.1) 19 ms 19 ms 19 ms 3 lilac-dmc.Berkeley.EDU (128.32.216.1) 39 ms 19 ms 19 ms 4 ccngw-ner-cc.Berkeley.EDU (128.32.136.23) 19 ms 39 ms 39 ms 5 ccn-nerif22.Berkeley.EDU (128.32.168.22) 20 ms 39 ms 39 ms 6 128.32.197.4 (128.32.197.4) 59 ms 119 ms 39 ms 7 131.119.2.5 (131.119.2.5) 59 ms 59 ms 39 ms 8 129.140.70.13 (129.140.70.13) 80 ms 79 ms 99 ms 9 129.140.71.6 (129.140.71.6) 139 ms 139 ms 159 ms 10 129.140.81.7 (129.140.81.7) 199 ms 180 ms 300 ms 11 129.140.72.17 (129.140.72.17) 300 ms 239 ms 239 ms 12 * * * 13 128.121.54.72 (128.121.54.72) 259 ms 499 ms 279 ms 14 * * * 15 * * * 16 * * * 17 * * * 18 ALLSPICE.LCS.MIT.EDU (18.26.0.115) 339 ms 279 ms 279 ms .Ed .Pp Note that the gateways 12, 14, 15, 16 & 17 hops away either don't send ICMP "time exceeded" messages or send them with a ttl too small to reach us. 14 \- 17 are running the MIT C Gateway code that doesn't send "time exceeded"s. God only knows what's going on with 12. .Pp The silent gateway 12 in the above may be the result of a bug in the 4.[23]BSD network code (and its derivatives): 4.x (x <= 3) sends an unreachable message using whatever ttl remains in the original datagram. Since, for gateways, the remaining ttl is zero, the ICMP "time exceeded" is guaranteed to not make it back to us. The behavior of this bug is slightly more interesting when it appears on the destination system: .Bd -literal -offset 4n 1 helios.ee.lbl.gov (128.3.112.1) 0 ms 0 ms 0 ms 2 lilac-dmc.Berkeley.EDU (128.32.216.1) 39 ms 19 ms 39 ms 3 lilac-dmc.Berkeley.EDU (128.32.216.1) 19 ms 39 ms 19 ms 4 ccngw-ner-cc.Berkeley.EDU (128.32.136.23) 39 ms 40 ms 19 ms 5 ccn-nerif35.Berkeley.EDU (128.32.168.35) 39 ms 39 ms 39 ms 6 csgw.Berkeley.EDU (128.32.133.254) 39 ms 59 ms 39 ms 7 * * * 8 * * * 9 * * * 10 * * * 11 * * * 12 * * * 13 rip.Berkeley.EDU (128.32.131.22) 59 ms ! 39 ms ! 39 ms ! .Ed .Pp Notice that there are 12 "gateways" (13 is the final destination) and exactly the last half of them are "missing". What's really happening is that rip (a Sun-3 running Sun OS3.5) is using the ttl from our arriving datagram as the ttl in its ICMP reply. So, the reply will time out on the return path (with no notice sent to anyone since ICMP's aren't sent for ICMP's) until we probe with a ttl that's at least twice the path length. I.e., rip is really only 7 hops away. A reply that returns with a ttl of 1 is a clue this problem exists. Traceroute prints a "!" after the time if the ttl is <= 1. Since vendors ship a lot of obsolete .Pf ( Tn DEC Ns \'s Ultrix, Sun 3.x) or non-standard .Pq Tn HP-UX software, expect to see this problem frequently and/or take care picking the target host of your probes. .Pp Other possible annotations after the time are: .Bl -hang -offset indent -width 12n .It Sy !H Host unreachable. .It Sy !N Network unreachable. .It Sy !P Protocol unreachable. .It Sy !S Source route failed. .It Sy !F\- Fragmentation needed. The RFC1191 Path MTU Discovery value is displayed. .It Sy !U Destination network unknown. .It Sy !W Destination host unknown. .It Sy !I Source host is isolated. .It Sy !A Communication with destination network administratively prohibited. .It Sy !Z Communication with destination host administratively prohibited. .It Sy !Q For this ToS the destination network is unreachable. .It Sy !T For this ToS the destination host is unreachable. .It Sy !X Communication administratively prohibited. .It Sy !V Host precedence violation. .It Sy !C Precedence cutoff in effect. .It Sy ! ICMP unreachable code . .El .Pp These are defined by RFC1812 (which supersedes RFC1716). If almost all the probes result in some kind of unreachable, .Nm will give up and exit. .Pp This program is intended for use in network testing, measurement and management. It should be used primarily for manual fault isolation. Because of the load it could impose on the network, it is unwise to use .Nm during normal operations or from automated scripts. .Sh SEE ALSO .Xr netstat 1 , .Xr ping 8 , .Xr ping6 8 , .Xr traceroute6 8 . .Sh AUTHORS Implemented by Van Jacobson from a suggestion by Steve Deering. Debugged by a cast of thousands with particularly cogent suggestions or fixes from C. Philip Wood, Tim Seaver and Ken Adelman. .Sh BUGS When using protocols other than UDP, functionality is reduced. In particular, the last packet will often appear to be lost, because even though it reaches the destination host, there's no way to know that because no ICMP message is sent back. In the TCP case, .Nm should listen for a RST from the destination host (or an intermediate router that's filtering packets), but this is not implemented yet. .Pp The AS number capability reports information that may sometimes be inaccurate due to discrepancies between the contents of the routing database server and the current state of the Internet. Index: projects/clang-trunk/contrib/traceroute/traceroute.c =================================================================== --- projects/clang-trunk/contrib/traceroute/traceroute.c (revision 283870) +++ projects/clang-trunk/contrib/traceroute/traceroute.c (revision 283871) @@ -1,1792 +1,1981 @@ /* * Copyright (c) 1988, 1989, 1991, 1994, 1995, 1996, 1997, 1998, 1999, 2000 * The Regents of the University of California. All rights reserved. * * Redistribution and use in source and binary forms, with or without * modification, are permitted provided that: (1) source code distributions * retain the above copyright notice and this paragraph in its entirety, (2) * distributions including binary code include the above copyright notice and * this paragraph in its entirety in the documentation or other materials * provided with the distribution, and (3) all advertising materials mentioning * features or use of this software display the following acknowledgement: * ``This product includes software developed by the University of California, * Lawrence Berkeley Laboratory and its contributors.'' Neither the name of * the University nor the names of its contributors may be used to endorse * or promote products derived from this software without specific prior * written permission. * THIS SOFTWARE IS PROVIDED ``AS IS'' AND WITHOUT ANY EXPRESS OR IMPLIED * WARRANTIES, INCLUDING, WITHOUT LIMITATION, THE IMPLIED WARRANTIES OF * MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE. */ #ifndef lint static const char copyright[] = "@(#) Copyright (c) 1988, 1989, 1991, 1994, 1995, 1996, 1997, 1998, 1999, 2000\n\ The Regents of the University of California. All rights reserved.\n"; #if 0 static const char rcsid[] = "@(#)$Id: traceroute.c,v 1.68 2000/12/14 08:04:33 leres Exp $ (LBL)"; #endif static const char rcsid[] = "$FreeBSD$"; #endif /* * traceroute host - trace the route ip packets follow going to "host". * * Attempt to trace the route an ip packet would follow to some * internet host. We find out intermediate hops by launching probe * packets with a small ttl (time to live) then listening for an * icmp "time exceeded" reply from a gateway. We start our probes * with a ttl of one and increase by one until we get an icmp "port * unreachable" (which means we got to "host") or hit a max (which * defaults to net.inet.ip.ttl hops & can be changed with the -m flag). * Three probes (change with -q flag) are sent at each ttl setting and * a line is printed showing the ttl, address of the gateway and * round trip time of each probe. If the probe answers come from * different gateways, the address of each responding system will * be printed. If there is no response within a 5 sec. timeout * interval (changed with the -w flag), a "*" is printed for that * probe. * * Probe packets are UDP format. We don't want the destination * host to process them so the destination port is set to an * unlikely value (if some clod on the destination is using that * value, it can be changed with the -p flag). * * A sample use might be: * * [yak 71]% traceroute nis.nsf.net. * traceroute to nis.nsf.net (35.1.1.48), 64 hops max, 56 byte packet * 1 helios.ee.lbl.gov (128.3.112.1) 19 ms 19 ms 0 ms * 2 lilac-dmc.Berkeley.EDU (128.32.216.1) 39 ms 39 ms 19 ms * 3 lilac-dmc.Berkeley.EDU (128.32.216.1) 39 ms 39 ms 19 ms * 4 ccngw-ner-cc.Berkeley.EDU (128.32.136.23) 39 ms 40 ms 39 ms * 5 ccn-nerif22.Berkeley.EDU (128.32.168.22) 39 ms 39 ms 39 ms * 6 128.32.197.4 (128.32.197.4) 40 ms 59 ms 59 ms * 7 131.119.2.5 (131.119.2.5) 59 ms 59 ms 59 ms * 8 129.140.70.13 (129.140.70.13) 99 ms 99 ms 80 ms * 9 129.140.71.6 (129.140.71.6) 139 ms 239 ms 319 ms * 10 129.140.81.7 (129.140.81.7) 220 ms 199 ms 199 ms * 11 nic.merit.edu (35.1.1.48) 239 ms 239 ms 239 ms * * Note that lines 2 & 3 are the same. This is due to a buggy * kernel on the 2nd hop system -- lbl-csam.arpa -- that forwards * packets with a zero ttl. * * A more interesting example is: * * [yak 72]% traceroute allspice.lcs.mit.edu. * traceroute to allspice.lcs.mit.edu (18.26.0.115), 64 hops max * 1 helios.ee.lbl.gov (128.3.112.1) 0 ms 0 ms 0 ms * 2 lilac-dmc.Berkeley.EDU (128.32.216.1) 19 ms 19 ms 19 ms * 3 lilac-dmc.Berkeley.EDU (128.32.216.1) 39 ms 19 ms 19 ms * 4 ccngw-ner-cc.Berkeley.EDU (128.32.136.23) 19 ms 39 ms 39 ms * 5 ccn-nerif22.Berkeley.EDU (128.32.168.22) 20 ms 39 ms 39 ms * 6 128.32.197.4 (128.32.197.4) 59 ms 119 ms 39 ms * 7 131.119.2.5 (131.119.2.5) 59 ms 59 ms 39 ms * 8 129.140.70.13 (129.140.70.13) 80 ms 79 ms 99 ms * 9 129.140.71.6 (129.140.71.6) 139 ms 139 ms 159 ms * 10 129.140.81.7 (129.140.81.7) 199 ms 180 ms 300 ms * 11 129.140.72.17 (129.140.72.17) 300 ms 239 ms 239 ms * 12 * * * * 13 128.121.54.72 (128.121.54.72) 259 ms 499 ms 279 ms * 14 * * * * 15 * * * * 16 * * * * 17 * * * * 18 ALLSPICE.LCS.MIT.EDU (18.26.0.115) 339 ms 279 ms 279 ms * * (I start to see why I'm having so much trouble with mail to * MIT.) Note that the gateways 12, 14, 15, 16 & 17 hops away * either don't send ICMP "time exceeded" messages or send them * with a ttl too small to reach us. 14 - 17 are running the * MIT C Gateway code that doesn't send "time exceeded"s. God * only knows what's going on with 12. * * The silent gateway 12 in the above may be the result of a bug in * the 4.[23]BSD network code (and its derivatives): 4.x (x <= 3) * sends an unreachable message using whatever ttl remains in the * original datagram. Since, for gateways, the remaining ttl is * zero, the icmp "time exceeded" is guaranteed to not make it back * to us. The behavior of this bug is slightly more interesting * when it appears on the destination system: * * 1 helios.ee.lbl.gov (128.3.112.1) 0 ms 0 ms 0 ms * 2 lilac-dmc.Berkeley.EDU (128.32.216.1) 39 ms 19 ms 39 ms * 3 lilac-dmc.Berkeley.EDU (128.32.216.1) 19 ms 39 ms 19 ms * 4 ccngw-ner-cc.Berkeley.EDU (128.32.136.23) 39 ms 40 ms 19 ms * 5 ccn-nerif35.Berkeley.EDU (128.32.168.35) 39 ms 39 ms 39 ms * 6 csgw.Berkeley.EDU (128.32.133.254) 39 ms 59 ms 39 ms * 7 * * * * 8 * * * * 9 * * * * 10 * * * * 11 * * * * 12 * * * * 13 rip.Berkeley.EDU (128.32.131.22) 59 ms ! 39 ms ! 39 ms ! * * Notice that there are 12 "gateways" (13 is the final * destination) and exactly the last half of them are "missing". * What's really happening is that rip (a Sun-3 running Sun OS3.5) * is using the ttl from our arriving datagram as the ttl in its * icmp reply. So, the reply will time out on the return path * (with no notice sent to anyone since icmp's aren't sent for * icmp's) until we probe with a ttl that's at least twice the path * length. I.e., rip is really only 7 hops away. A reply that * returns with a ttl of 1 is a clue this problem exists. * Traceroute prints a "!" after the time if the ttl is <= 1. * Since vendors ship a lot of obsolete (DEC's Ultrix, Sun 3.x) or * non-standard (HPUX) software, expect to see this problem * frequently and/or take care picking the target host of your * probes. * * Other possible annotations after the time are !H, !N, !P (got a host, * network or protocol unreachable, respectively), !S or !F (source * route failed or fragmentation needed -- neither of these should * ever occur and the associated gateway is busted if you see one). If * almost all the probes result in some kind of unreachable, traceroute * will give up and exit. * * Notes * ----- * This program must be run by root or be setuid. (I suggest that * you *don't* make it setuid -- casual use could result in a lot * of unnecessary traffic on our poor, congested nets.) * * This program requires a kernel mod that does not appear in any * system available from Berkeley: A raw ip socket using proto * IPPROTO_RAW must interpret the data sent as an ip datagram (as * opposed to data to be wrapped in a ip datagram). See the README * file that came with the source to this program for a description * of the mods I made to /sys/netinet/raw_ip.c. Your mileage may * vary. But, again, ANY 4.x (x < 4) BSD KERNEL WILL HAVE TO BE * MODIFIED TO RUN THIS PROGRAM. * * The udp port usage may appear bizarre (well, ok, it is bizarre). * The problem is that an icmp message only contains 8 bytes of * data from the original datagram. 8 bytes is the size of a udp * header so, if we want to associate replies with the original * datagram, the necessary information must be encoded into the * udp header (the ip id could be used but there's no way to * interlock with the kernel's assignment of ip id's and, anyway, * it would have taken a lot more kernel hacking to allow this * code to set the ip id). So, to allow two or more users to * use traceroute simultaneously, we use this task's pid as the * source port (the high bit is set to move the port number out * of the "likely" range). To keep track of which probe is being * replied to (so times and/or hop counts don't get confused by a * reply that was delayed in transit), we increment the destination * port number before each probe. * * Don't use this as a coding example. I was trying to find a * routing problem and this code sort-of popped out after 48 hours * without sleep. I was amazed it ever compiled, much less ran. * * I stole the idea for this program from Steve Deering. Since * the first release, I've learned that had I attended the right * IETF working group meetings, I also could have stolen it from Guy * Almes or Matt Mathis. I don't know (or care) who came up with * the idea first. I envy the originators' perspicacity and I'm * glad they didn't keep the idea a secret. * * Tim Seaver, Ken Adelman and C. Philip Wood provided bug fixes and/or * enhancements to the original distribution. * * I've hacked up a round-trip-route version of this that works by * sending a loose-source-routed udp datagram through the destination * back to yourself. Unfortunately, SO many gateways botch source * routing, the thing is almost worthless. Maybe one day... * * -- Van Jacobson (van@ee.lbl.gov) * Tue Dec 20 03:50:13 PST 1988 */ #include #include #include #ifdef HAVE_SYS_SELECT_H #include #endif #include #ifdef HAVE_SYS_SYSCTL_H #include #endif #include #include #include #include #include #include +#include #include #include #include #include #ifdef IPSEC #include #include /* XXX */ #endif /* IPSEC */ #include #include #include #include #ifdef HAVE_MALLOC_H #include #endif #include #include #include #include #include #include /* rfc1716 */ #ifndef ICMP_UNREACH_FILTER_PROHIB #define ICMP_UNREACH_FILTER_PROHIB 13 /* admin prohibited filter */ #endif #ifndef ICMP_UNREACH_HOST_PRECEDENCE #define ICMP_UNREACH_HOST_PRECEDENCE 14 /* host precedence violation */ #endif #ifndef ICMP_UNREACH_PRECEDENCE_CUTOFF #define ICMP_UNREACH_PRECEDENCE_CUTOFF 15 /* precedence cutoff */ #endif #include "findsaddr.h" #include "ifaddrlist.h" #include "as.h" #include "traceroute.h" /* Maximum number of gateways (include room for one noop) */ #define NGATEWAYS ((int)((MAX_IPOPTLEN - IPOPT_MINOFF - 1) / sizeof(u_int32_t))) #ifndef MAXHOSTNAMELEN #define MAXHOSTNAMELEN 64 #endif #define Fprintf (void)fprintf #define Printf (void)printf /* What a GRE packet header looks like */ struct grehdr { u_int16_t flags; u_int16_t proto; u_int16_t length; /* PPTP version of these fields */ u_int16_t callId; }; #ifndef IPPROTO_GRE #define IPPROTO_GRE 47 #endif /* For GRE, we prepare what looks like a PPTP packet */ #define GRE_PPTP_PROTO 0x880b /* Host name and address list */ struct hostinfo { char *name; int n; u_int32_t *addrs; }; /* Data section of the probe packet */ struct outdata { u_char seq; /* sequence number of this packet */ u_char ttl; /* ttl packet left with */ struct timeval tv; /* time packet left */ }; #ifndef HAVE_ICMP_NEXTMTU /* Path MTU Discovery (RFC1191) */ struct my_pmtu { u_short ipm_void; u_short ipm_nextmtu; }; #endif u_char packet[512]; /* last inbound (icmp) packet */ struct ip *outip; /* last output ip packet */ u_char *outp; /* last output inner protocol packet */ struct ip *hip = NULL; /* Quoted IP header */ int hiplen = 0; /* loose source route gateway list (including room for final destination) */ u_int32_t gwlist[NGATEWAYS + 1]; int s; /* receive (icmp) socket file descriptor */ int sndsock; /* send (udp) socket file descriptor */ struct sockaddr whereto; /* Who to try to reach */ struct sockaddr wherefrom; /* Who we are */ int packlen; /* total length of packet */ int protlen; /* length of protocol part of packet */ int minpacket; /* min ip packet size */ int maxpacket = 32 * 1024; /* max ip packet size */ int pmtu; /* Path MTU Discovery (RFC1191) */ u_int pausemsecs; char *prog; char *source; char *hostname; char *device; static const char devnull[] = "/dev/null"; int nprobes = -1; int max_ttl; int first_ttl = 1; u_short ident; u_short port; /* protocol specific base "port" */ int options; /* socket options */ int verbose; int waittime = 5; /* time to wait for response (in seconds) */ int nflag; /* print addresses numerically */ int as_path; /* print as numbers for each hop */ char *as_server = NULL; void *asn; #ifdef CANT_HACK_IPCKSUM int doipcksum = 0; /* don't calculate ip checksums by default */ #else int doipcksum = 1; /* calculate ip checksums by default */ #endif int optlen; /* length of ip options */ int fixedPort = 0; /* Use fixed destination port for TCP and UDP */ int printdiff = 0; /* Print the difference between sent and quoted */ extern int optind; extern int opterr; extern char *optarg; /* Forwards */ double deltaT(struct timeval *, struct timeval *); void freehostinfo(struct hostinfo *); void getaddr(u_int32_t *, char *); struct hostinfo *gethostinfo(char *); u_short in_cksum(u_short *, int); +u_int32_t sctp_crc32c(const void *, u_int32_t); char *inetname(struct in_addr); int main(int, char **); -u_short p_cksum(struct ip *, u_short *, int); +u_short p_cksum(struct ip *, u_short *, int, int); int packet_ok(u_char *, int, struct sockaddr_in *, int); char *pr_type(u_char); void print(u_char *, int, struct sockaddr_in *); #ifdef IPSEC int setpolicy __P((int so, char *policy)); #endif void send_probe(int, int); struct outproto *setproto(char *); int str2val(const char *, const char *, int, int); void tvsub(struct timeval *, struct timeval *); void usage(void); int wait_for_reply(int, struct sockaddr_in *, const struct timeval *); void pkt_compare(const u_char *, int, const u_char *, int); #ifndef HAVE_USLEEP int usleep(u_int); #endif void udp_prep(struct outdata *); int udp_check(const u_char *, int); +void udplite_prep(struct outdata *); +int udplite_check(const u_char *, int); void tcp_prep(struct outdata *); int tcp_check(const u_char *, int); +void sctp_prep(struct outdata *); +int sctp_check(const u_char *, int); void gre_prep(struct outdata *); int gre_check(const u_char *, int); void gen_prep(struct outdata *); int gen_check(const u_char *, int); void icmp_prep(struct outdata *); int icmp_check(const u_char *, int); /* Descriptor structure for each outgoing protocol we support */ struct outproto { char *name; /* name of protocol */ const char *key; /* An ascii key for the bytes of the header */ u_char num; /* IP protocol number */ u_short hdrlen; /* max size of protocol header */ u_short port; /* default base protocol-specific "port" */ void (*prepare)(struct outdata *); /* finish preparing an outgoing packet */ int (*check)(const u_char *, int); /* check an incoming packet */ }; /* List of supported protocols. The first one is the default. The last one is the handler for generic protocols not explicitly listed. */ struct outproto protos[] = { { "udp", "spt dpt len sum", IPPROTO_UDP, sizeof(struct udphdr), 32768 + 666, udp_prep, udp_check }, { + "udplite", + "spt dpt cov sum", + IPPROTO_UDPLITE, + sizeof(struct udphdr), + 32768 + 666, + udplite_prep, + udplite_check + }, + { "tcp", "spt dpt seq ack xxflwin sum urp", IPPROTO_TCP, sizeof(struct tcphdr), 32768 + 666, tcp_prep, tcp_check }, { + "sctp", + "spt dpt vtag crc tyfllen tyfllen ", + IPPROTO_SCTP, + sizeof(struct sctphdr), + 32768 + 666, + sctp_prep, + sctp_check + }, + { "gre", "flg pro len clid", IPPROTO_GRE, sizeof(struct grehdr), GRE_PPTP_PROTO, gre_prep, gre_check }, { "icmp", "typ cod sum ", IPPROTO_ICMP, sizeof(struct icmp), 0, icmp_prep, icmp_check }, { NULL, - NULL, + "", 0, 2 * sizeof(u_short), 0, gen_prep, gen_check }, }; struct outproto *proto = &protos[0]; const char *ip_hdr_key = "vhtslen id off tlprsum srcip dstip opts"; int main(int argc, char **argv) { register int op, code, n; register char *cp; register const char *err; register u_int32_t *ap; register struct sockaddr_in *from = (struct sockaddr_in *)&wherefrom; register struct sockaddr_in *to = (struct sockaddr_in *)&whereto; register struct hostinfo *hi; int on = 1; register struct protoent *pe; register int ttl, probe, i; register int seq = 0; int tos = 0, settos = 0; register int lsrr = 0; register u_short off = 0; struct ifaddrlist *al; char errbuf[132]; int requestPort = -1; int sump = 0; int sockerrno; /* Insure the socket fds won't be 0, 1 or 2 */ if (open(devnull, O_RDONLY) < 0 || open(devnull, O_RDONLY) < 0 || open(devnull, O_RDONLY) < 0) { Fprintf(stderr, "%s: open \"%s\": %s\n", prog, devnull, strerror(errno)); exit(1); } /* * Do the setuid-required stuff first, then lose priveleges ASAP. * Do error checking for these two calls where they appeared in * the original code. */ cp = "icmp"; pe = getprotobyname(cp); if (pe) { if ((s = socket(AF_INET, SOCK_RAW, pe->p_proto)) < 0) sockerrno = errno; else if ((sndsock = socket(AF_INET, SOCK_RAW, IPPROTO_RAW)) < 0) sockerrno = errno; } if (setuid(getuid()) != 0) { perror("setuid()"); exit(1); } #ifdef IPCTL_DEFTTL { int mib[4] = { CTL_NET, PF_INET, IPPROTO_IP, IPCTL_DEFTTL }; size_t sz = sizeof(max_ttl); if (sysctl(mib, 4, &max_ttl, &sz, NULL, 0) == -1) { perror("sysctl(net.inet.ip.ttl)"); exit(1); } } #else max_ttl = 30; #endif if (argv[0] == NULL) prog = "traceroute"; else if ((cp = strrchr(argv[0], '/')) != NULL) prog = cp + 1; else prog = argv[0]; opterr = 0; while ((op = getopt(argc, argv, "aA:edDFInrSvxf:g:i:M:m:P:p:q:s:t:w:z:")) != EOF) switch (op) { case 'a': as_path = 1; break; - + case 'A': as_path = 1; as_server = optarg; break; - + case 'd': options |= SO_DEBUG; break; case 'D': printdiff = 1; break; case 'e': fixedPort = 1; break; case 'f': case 'M': /* FreeBSD compat. */ first_ttl = str2val(optarg, "first ttl", 1, 255); break; case 'F': off = IP_DF; break; case 'g': if (lsrr >= NGATEWAYS) { Fprintf(stderr, "%s: No more than %d gateways\n", prog, NGATEWAYS); exit(1); } getaddr(gwlist + lsrr, optarg); ++lsrr; break; case 'i': device = optarg; break; case 'I': proto = setproto("icmp"); break; case 'm': max_ttl = str2val(optarg, "max ttl", 1, 255); break; case 'n': ++nflag; break; case 'P': proto = setproto(optarg); break; case 'p': requestPort = (u_short)str2val(optarg, "port", 1, (1 << 16) - 1); break; case 'q': nprobes = str2val(optarg, "nprobes", 1, -1); break; case 'r': options |= SO_DONTROUTE; break; case 's': /* * set the ip source address of the outbound * probe (e.g., on a multi-homed host). */ source = optarg; break; case 'S': sump = 1; break; case 't': tos = str2val(optarg, "tos", 0, 255); ++settos; break; case 'v': ++verbose; break; case 'x': doipcksum = (doipcksum == 0); break; case 'w': waittime = str2val(optarg, "wait time", 1, 24 * 60 * 60); break; case 'z': pausemsecs = str2val(optarg, "pause msecs", 0, 60 * 60 * 1000); break; default: usage(); } /* Set requested port, if any, else default for this protocol */ port = (requestPort != -1) ? requestPort : proto->port; if (nprobes == -1) nprobes = printdiff ? 1 : 3; if (first_ttl > max_ttl) { Fprintf(stderr, "%s: first ttl (%d) may not be greater than max ttl (%d)\n", prog, first_ttl, max_ttl); exit(1); } if (!doipcksum) Fprintf(stderr, "%s: Warning: ip checksums disabled\n", prog); if (lsrr > 0) optlen = (lsrr + 1) * sizeof(gwlist[0]); - minpacket = sizeof(*outip) + proto->hdrlen + sizeof(struct outdata) + optlen; - packlen = minpacket; /* minimum sized packet */ + minpacket = sizeof(*outip) + proto->hdrlen + optlen; + if (minpacket > 40) + packlen = minpacket; + else + packlen = 40; /* Process destination and optional packet size */ switch (argc - optind) { case 2: packlen = str2val(argv[optind + 1], "packet length", minpacket, maxpacket); /* Fall through */ case 1: hostname = argv[optind]; hi = gethostinfo(hostname); setsin(to, hi->addrs[0]); if (hi->n > 1) Fprintf(stderr, "%s: Warning: %s has multiple addresses; using %s\n", prog, hostname, inet_ntoa(to->sin_addr)); hostname = hi->name; hi->name = NULL; freehostinfo(hi); break; default: usage(); } #ifdef HAVE_SETLINEBUF setlinebuf (stdout); #else setvbuf(stdout, NULL, _IOLBF, 0); #endif protlen = packlen - sizeof(*outip) - optlen; + if ((proto->num == IPPROTO_SCTP) && (packlen & 3)) { + Fprintf(stderr, "%s: packet length must be a multiple of 4\n", + prog); + exit(1); + } outip = (struct ip *)malloc((unsigned)packlen); if (outip == NULL) { Fprintf(stderr, "%s: malloc: %s\n", prog, strerror(errno)); exit(1); } memset((char *)outip, 0, packlen); outip->ip_v = IPVERSION; if (settos) outip->ip_tos = tos; #ifdef BYTESWAP_IP_HDR outip->ip_len = htons(packlen); outip->ip_off = htons(off); #else outip->ip_len = packlen; outip->ip_off = off; #endif outip->ip_p = proto->num; outp = (u_char *)(outip + 1); #ifdef HAVE_RAW_OPTIONS if (lsrr > 0) { register u_char *optlist; optlist = outp; outp += optlen; /* final hop */ gwlist[lsrr] = to->sin_addr.s_addr; outip->ip_dst.s_addr = gwlist[0]; /* force 4 byte alignment */ optlist[0] = IPOPT_NOP; /* loose source route option */ optlist[1] = IPOPT_LSRR; i = lsrr * sizeof(gwlist[0]); optlist[2] = i + 3; /* Pointer to LSRR addresses */ optlist[3] = IPOPT_MINOFF; memcpy(optlist + 4, gwlist + 1, i); } else #endif outip->ip_dst = to->sin_addr; outip->ip_hl = (outp - (u_char *)outip) >> 2; ident = (getpid() & 0xffff) | 0x8000; if (pe == NULL) { Fprintf(stderr, "%s: unknown protocol %s\n", prog, cp); exit(1); } if (s < 0) { errno = sockerrno; Fprintf(stderr, "%s: icmp socket: %s\n", prog, strerror(errno)); exit(1); } if (options & SO_DEBUG) (void)setsockopt(s, SOL_SOCKET, SO_DEBUG, (char *)&on, sizeof(on)); if (options & SO_DONTROUTE) (void)setsockopt(s, SOL_SOCKET, SO_DONTROUTE, (char *)&on, sizeof(on)); #if defined(IPSEC) && defined(IPSEC_POLICY_IPSEC) if (setpolicy(s, "in bypass") < 0) errx(1, "%s", ipsec_strerror()); if (setpolicy(s, "out bypass") < 0) errx(1, "%s", ipsec_strerror()); #endif /* defined(IPSEC) && defined(IPSEC_POLICY_IPSEC) */ if (sndsock < 0) { errno = sockerrno; Fprintf(stderr, "%s: raw socket: %s\n", prog, strerror(errno)); exit(1); } #if defined(IP_OPTIONS) && !defined(HAVE_RAW_OPTIONS) if (lsrr > 0) { u_char optlist[MAX_IPOPTLEN]; cp = "ip"; if ((pe = getprotobyname(cp)) == NULL) { Fprintf(stderr, "%s: unknown protocol %s\n", prog, cp); exit(1); } /* final hop */ gwlist[lsrr] = to->sin_addr.s_addr; ++lsrr; /* force 4 byte alignment */ optlist[0] = IPOPT_NOP; /* loose source route option */ optlist[1] = IPOPT_LSRR; i = lsrr * sizeof(gwlist[0]); optlist[2] = i + 3; /* Pointer to LSRR addresses */ optlist[3] = IPOPT_MINOFF; memcpy(optlist + 4, gwlist, i); if ((setsockopt(sndsock, pe->p_proto, IP_OPTIONS, (char *)optlist, i + sizeof(gwlist[0]))) < 0) { Fprintf(stderr, "%s: IP_OPTIONS: %s\n", prog, strerror(errno)); exit(1); } } #endif #ifdef SO_SNDBUF if (setsockopt(sndsock, SOL_SOCKET, SO_SNDBUF, (char *)&packlen, sizeof(packlen)) < 0) { Fprintf(stderr, "%s: SO_SNDBUF: %s\n", prog, strerror(errno)); exit(1); } #endif #ifdef IP_HDRINCL if (setsockopt(sndsock, IPPROTO_IP, IP_HDRINCL, (char *)&on, sizeof(on)) < 0) { Fprintf(stderr, "%s: IP_HDRINCL: %s\n", prog, strerror(errno)); exit(1); } #else #ifdef IP_TOS if (settos && setsockopt(sndsock, IPPROTO_IP, IP_TOS, (char *)&tos, sizeof(tos)) < 0) { Fprintf(stderr, "%s: setsockopt tos %d: %s\n", prog, tos, strerror(errno)); exit(1); } #endif #endif if (options & SO_DEBUG) (void)setsockopt(sndsock, SOL_SOCKET, SO_DEBUG, (char *)&on, sizeof(on)); if (options & SO_DONTROUTE) (void)setsockopt(sndsock, SOL_SOCKET, SO_DONTROUTE, (char *)&on, sizeof(on)); /* Get the interface address list */ n = ifaddrlist(&al, errbuf); if (n < 0) { Fprintf(stderr, "%s: ifaddrlist: %s\n", prog, errbuf); exit(1); } if (n == 0) { Fprintf(stderr, "%s: Can't find any network interfaces\n", prog); exit(1); } /* Look for a specific device */ if (device != NULL) { for (i = n; i > 0; --i, ++al) if (strcmp(device, al->device) == 0) break; if (i <= 0) { Fprintf(stderr, "%s: Can't find interface %.32s\n", prog, device); exit(1); } } /* Determine our source address */ if (source == NULL) { /* * If a device was specified, use the interface address. * Otherwise, try to determine our source address. */ if (device != NULL) setsin(from, al->addr); else if ((err = findsaddr(to, from)) != NULL) { Fprintf(stderr, "%s: findsaddr: %s\n", prog, err); exit(1); } } else { hi = gethostinfo(source); source = hi->name; hi->name = NULL; /* * If the device was specified make sure it * corresponds to the source address specified. * Otherwise, use the first address (and warn if * there are more than one). */ if (device != NULL) { for (i = hi->n, ap = hi->addrs; i > 0; --i, ++ap) if (*ap == al->addr) break; if (i <= 0) { Fprintf(stderr, "%s: %s is not on interface %.32s\n", prog, source, device); exit(1); } setsin(from, *ap); } else { setsin(from, hi->addrs[0]); if (hi->n > 1) Fprintf(stderr, "%s: Warning: %s has multiple addresses; using %s\n", prog, source, inet_ntoa(from->sin_addr)); } freehostinfo(hi); } outip->ip_src = from->sin_addr; /* Check the source address (-s), if any, is valid */ if (bind(sndsock, (struct sockaddr *)from, sizeof(*from)) < 0) { Fprintf(stderr, "%s: bind: %s\n", prog, strerror(errno)); exit (1); } if (as_path) { asn = as_setup(as_server); if (asn == NULL) { Fprintf(stderr, "%s: as_setup failed, AS# lookups" " disabled\n", prog); (void)fflush(stderr); as_path = 0; } } - + #if defined(IPSEC) && defined(IPSEC_POLICY_IPSEC) if (setpolicy(sndsock, "in bypass") < 0) errx(1, "%s", ipsec_strerror()); if (setpolicy(sndsock, "out bypass") < 0) errx(1, "%s", ipsec_strerror()); #endif /* defined(IPSEC) && defined(IPSEC_POLICY_IPSEC) */ Fprintf(stderr, "%s to %s (%s)", prog, hostname, inet_ntoa(to->sin_addr)); if (source) Fprintf(stderr, " from %s", source); Fprintf(stderr, ", %d hops max, %d byte packets\n", max_ttl, packlen); (void)fflush(stderr); for (ttl = first_ttl; ttl <= max_ttl; ++ttl) { u_int32_t lastaddr = 0; int gotlastaddr = 0; int got_there = 0; int unreachable = 0; int sentfirst = 0; int loss; Printf("%2d ", ttl); for (probe = 0, loss = 0; probe < nprobes; ++probe) { register int cc; struct timeval t1, t2; register struct ip *ip; struct outdata outdata; if (sentfirst && pausemsecs > 0) usleep(pausemsecs * 1000); /* Prepare outgoing data */ outdata.seq = ++seq; outdata.ttl = ttl; /* Avoid alignment problems by copying bytewise: */ (void)gettimeofday(&t1, NULL); memcpy(&outdata.tv, &t1, sizeof(outdata.tv)); /* Finalize and send packet */ (*proto->prepare)(&outdata); send_probe(seq, ttl); ++sentfirst; /* Wait for a reply */ while ((cc = wait_for_reply(s, from, &t1)) != 0) { double T; int precis; (void)gettimeofday(&t2, NULL); i = packet_ok(packet, cc, from, seq); /* Skip short packet */ if (i == 0) continue; if (!gotlastaddr || from->sin_addr.s_addr != lastaddr) { if (gotlastaddr) printf("\n "); print(packet, cc, from); lastaddr = from->sin_addr.s_addr; ++gotlastaddr; } T = deltaT(&t1, &t2); #ifdef SANE_PRECISION if (T >= 1000.0) precis = 0; else if (T >= 100.0) precis = 1; else if (T >= 10.0) precis = 2; else #endif precis = 3; Printf(" %.*f ms", precis, T); if (printdiff) { Printf("\n"); Printf("%*.*s%s\n", -(outip->ip_hl << 3), outip->ip_hl << 3, ip_hdr_key, proto->key); pkt_compare((void *)outip, packlen, (void *)hip, hiplen); } if (i == -2) { #ifndef ARCHAIC ip = (struct ip *)packet; if (ip->ip_ttl <= 1) Printf(" !"); #endif ++got_there; break; } /* time exceeded in transit */ if (i == -1) break; code = i - 1; switch (code) { case ICMP_UNREACH_PORT: #ifndef ARCHAIC ip = (struct ip *)packet; if (ip->ip_ttl <= 1) Printf(" !"); #endif ++got_there; break; case ICMP_UNREACH_NET: ++unreachable; Printf(" !N"); break; case ICMP_UNREACH_HOST: ++unreachable; Printf(" !H"); break; case ICMP_UNREACH_PROTOCOL: ++got_there; Printf(" !P"); break; case ICMP_UNREACH_NEEDFRAG: ++unreachable; Printf(" !F-%d", pmtu); break; case ICMP_UNREACH_SRCFAIL: ++unreachable; Printf(" !S"); break; case ICMP_UNREACH_NET_UNKNOWN: ++unreachable; Printf(" !U"); break; case ICMP_UNREACH_HOST_UNKNOWN: ++unreachable; Printf(" !W"); break; case ICMP_UNREACH_ISOLATED: ++unreachable; Printf(" !I"); break; case ICMP_UNREACH_NET_PROHIB: ++unreachable; Printf(" !A"); break; case ICMP_UNREACH_HOST_PROHIB: ++unreachable; Printf(" !Z"); break; case ICMP_UNREACH_TOSNET: ++unreachable; Printf(" !Q"); break; case ICMP_UNREACH_TOSHOST: ++unreachable; Printf(" !T"); break; case ICMP_UNREACH_FILTER_PROHIB: ++unreachable; Printf(" !X"); break; case ICMP_UNREACH_HOST_PRECEDENCE: ++unreachable; Printf(" !V"); break; case ICMP_UNREACH_PRECEDENCE_CUTOFF: ++unreachable; Printf(" !C"); break; default: ++unreachable; Printf(" !<%d>", code); break; } break; } if (cc == 0) { loss++; Printf(" *"); } (void)fflush(stdout); } if (sump) { Printf(" (%d%% loss)", (loss * 100) / nprobes); } putchar('\n'); if (got_there || (unreachable > 0 && unreachable >= nprobes - 1)) break; } if (as_path) as_shutdown(asn); exit(0); } int wait_for_reply(register int sock, register struct sockaddr_in *fromp, register const struct timeval *tp) { fd_set *fdsp; size_t nfds; struct timeval now, wait; register int cc = 0; register int error; int fromlen = sizeof(*fromp); nfds = howmany(sock + 1, NFDBITS); if ((fdsp = malloc(nfds * sizeof(fd_mask))) == NULL) err(1, "malloc"); memset(fdsp, 0, nfds * sizeof(fd_mask)); FD_SET(sock, fdsp); wait.tv_sec = tp->tv_sec + waittime; wait.tv_usec = tp->tv_usec; (void)gettimeofday(&now, NULL); tvsub(&wait, &now); if (wait.tv_sec < 0) { wait.tv_sec = 0; wait.tv_usec = 1; } error = select(sock + 1, fdsp, NULL, NULL, &wait); if (error == -1 && errno == EINVAL) { Fprintf(stderr, "%s: botched select() args\n", prog); exit(1); } if (error > 0) cc = recvfrom(sock, (char *)packet, sizeof(packet), 0, (struct sockaddr *)fromp, &fromlen); free(fdsp); return(cc); } void send_probe(int seq, int ttl) { register int cc; outip->ip_ttl = ttl; outip->ip_id = htons(ident + seq); /* XXX undocumented debugging hack */ if (verbose > 1) { register const u_short *sp; register int nshorts, i; sp = (u_short *)outip; nshorts = (u_int)packlen / sizeof(u_short); i = 0; Printf("[ %d bytes", packlen); while (--nshorts >= 0) { if ((i++ % 8) == 0) Printf("\n\t"); Printf(" %04x", ntohs(*sp++)); } if (packlen & 1) { if ((i % 8) == 0) Printf("\n\t"); Printf(" %02x", *(u_char *)sp); } Printf("]\n"); } #if !defined(IP_HDRINCL) && defined(IP_TTL) if (setsockopt(sndsock, IPPROTO_IP, IP_TTL, (char *)&ttl, sizeof(ttl)) < 0) { Fprintf(stderr, "%s: setsockopt ttl %d: %s\n", prog, ttl, strerror(errno)); exit(1); } #endif cc = sendto(sndsock, (char *)outip, packlen, 0, &whereto, sizeof(whereto)); if (cc < 0 || cc != packlen) { if (cc < 0) Fprintf(stderr, "%s: sendto: %s\n", prog, strerror(errno)); Printf("%s: wrote %s %d chars, ret=%d\n", prog, hostname, packlen, cc); (void)fflush(stdout); } } #if defined(IPSEC) && defined(IPSEC_POLICY_IPSEC) int setpolicy(so, policy) int so; char *policy; { char *buf; buf = ipsec_set_policy(policy, strlen(policy)); if (buf == NULL) { warnx("%s", ipsec_strerror()); return -1; } (void)setsockopt(so, IPPROTO_IP, IP_IPSEC_POLICY, buf, ipsec_get_policylen(buf)); free(buf); return 0; } #endif double deltaT(struct timeval *t1p, struct timeval *t2p) { register double dt; dt = (double)(t2p->tv_sec - t1p->tv_sec) * 1000.0 + (double)(t2p->tv_usec - t1p->tv_usec) / 1000.0; return (dt); } /* * Convert an ICMP "type" field to a printable string. */ char * pr_type(register u_char t) { static char *ttab[] = { "Echo Reply", "ICMP 1", "ICMP 2", "Dest Unreachable", "Source Quench", "Redirect", "ICMP 6", "ICMP 7", "Echo", "ICMP 9", "ICMP 10", "Time Exceeded", "Param Problem", "Timestamp", "Timestamp Reply", "Info Request", "Info Reply" }; if (t > 16) return("OUT-OF-RANGE"); return(ttab[t]); } int packet_ok(register u_char *buf, int cc, register struct sockaddr_in *from, register int seq) { register struct icmp *icp; register u_char type, code; register int hlen; #ifndef ARCHAIC register struct ip *ip; ip = (struct ip *) buf; hlen = ip->ip_hl << 2; if (cc < hlen + ICMP_MINLEN) { if (verbose) Printf("packet too short (%d bytes) from %s\n", cc, inet_ntoa(from->sin_addr)); return (0); } cc -= hlen; icp = (struct icmp *)(buf + hlen); #else icp = (struct icmp *)buf; #endif type = icp->icmp_type; code = icp->icmp_code; /* Path MTU Discovery (RFC1191) */ if (code != ICMP_UNREACH_NEEDFRAG) pmtu = 0; else { #ifdef HAVE_ICMP_NEXTMTU pmtu = ntohs(icp->icmp_nextmtu); #else pmtu = ntohs(((struct my_pmtu *)&icp->icmp_void)->ipm_nextmtu); #endif } if (type == ICMP_ECHOREPLY && proto->num == IPPROTO_ICMP && (*proto->check)((u_char *)icp, (u_char)seq)) return -2; if ((type == ICMP_TIMXCEED && code == ICMP_TIMXCEED_INTRANS) || type == ICMP_UNREACH) { u_char *inner; hip = &icp->icmp_ip; hiplen = ((u_char *)icp + cc) - (u_char *)hip; hlen = hip->ip_hl << 2; inner = (u_char *)((u_char *)hip + hlen); - if (hlen + 12 <= cc + if (hlen + 16 <= cc && hip->ip_p == proto->num && (*proto->check)(inner, (u_char)seq)) return (type == ICMP_TIMXCEED ? -1 : code + 1); } #ifndef ARCHAIC if (verbose) { register int i; u_int32_t *lp = (u_int32_t *)&icp->icmp_ip; Printf("\n%d bytes from %s to ", cc, inet_ntoa(from->sin_addr)); Printf("%s: icmp type %d (%s) code %d\n", inet_ntoa(ip->ip_dst), type, pr_type(type), icp->icmp_code); - for (i = 4; i < cc ; i += sizeof(*lp)) - Printf("%2d: x%8.8x\n", i, *lp++); + for (i = 4; i <= cc - ICMP_MINLEN; i += sizeof(*lp)) + Printf("%2d: %8.8x\n", i, ntohl(*lp++)); } #endif return(0); } void icmp_prep(struct outdata *outdata) { struct icmp *const icmpheader = (struct icmp *) outp; icmpheader->icmp_type = ICMP_ECHO; icmpheader->icmp_id = htons(ident); icmpheader->icmp_seq = htons(outdata->seq); icmpheader->icmp_cksum = 0; icmpheader->icmp_cksum = in_cksum((u_short *)icmpheader, protlen); if (icmpheader->icmp_cksum == 0) icmpheader->icmp_cksum = 0xffff; } int icmp_check(const u_char *data, int seq) { struct icmp *const icmpheader = (struct icmp *) data; return (icmpheader->icmp_id == htons(ident) && icmpheader->icmp_seq == htons(seq)); } void udp_prep(struct outdata *outdata) { struct udphdr *const outudp = (struct udphdr *) outp; outudp->uh_sport = htons(ident + (fixedPort ? outdata->seq : 0)); outudp->uh_dport = htons(port + (fixedPort ? 0 : outdata->seq)); outudp->uh_ulen = htons((u_short)protlen); outudp->uh_sum = 0; if (doipcksum) { - u_short sum = p_cksum(outip, (u_short*)outudp, protlen); + u_short sum = p_cksum(outip, (u_short*)outudp, protlen, protlen); outudp->uh_sum = (sum) ? sum : 0xffff; } return; } int udp_check(const u_char *data, int seq) { struct udphdr *const udp = (struct udphdr *) data; return (ntohs(udp->uh_sport) == ident + (fixedPort ? seq : 0) && ntohs(udp->uh_dport) == port + (fixedPort ? 0 : seq)); } void +udplite_prep(struct outdata *outdata) +{ + struct udphdr *const outudp = (struct udphdr *) outp; + + outudp->uh_sport = htons(ident + (fixedPort ? outdata->seq : 0)); + outudp->uh_dport = htons(port + (fixedPort ? 0 : outdata->seq)); + outudp->uh_ulen = htons(8); + outudp->uh_sum = 0; + if (doipcksum) { + u_short sum = p_cksum(outip, (u_short*)outudp, protlen, 8); + outudp->uh_sum = (sum) ? sum : 0xffff; + } + + return; +} + +int +udplite_check(const u_char *data, int seq) +{ + struct udphdr *const udp = (struct udphdr *) data; + + return (ntohs(udp->uh_sport) == ident + (fixedPort ? seq : 0) && + ntohs(udp->uh_dport) == port + (fixedPort ? 0 : seq)); +} + +void tcp_prep(struct outdata *outdata) { struct tcphdr *const tcp = (struct tcphdr *) outp; tcp->th_sport = htons(ident); tcp->th_dport = htons(port + (fixedPort ? 0 : outdata->seq)); tcp->th_seq = (tcp->th_sport << 16) | tcp->th_dport; tcp->th_ack = 0; tcp->th_off = 5; tcp->th_flags = TH_SYN; tcp->th_sum = 0; - if (doipcksum) { - u_short sum = p_cksum(outip, (u_short*)tcp, protlen); - tcp->th_sum = (sum) ? sum : 0xffff; - } + if (doipcksum) + tcp->th_sum = p_cksum(outip, (u_short*)tcp, protlen, protlen); } int tcp_check(const u_char *data, int seq) { struct tcphdr *const tcp = (struct tcphdr *) data; return (ntohs(tcp->th_sport) == ident && ntohs(tcp->th_dport) == port + (fixedPort ? 0 : seq) && tcp->th_seq == (tcp_seq)((tcp->th_sport << 16) | tcp->th_dport)); } void +sctp_prep(struct outdata *outdata) +{ + struct sctphdr *const sctp = (struct sctphdr *) outp; + struct sctp_chunkhdr *chk; + + sctp->src_port = htons(ident); + sctp->dest_port = htons(port + (fixedPort ? 0 : outdata->seq)); + sctp->v_tag = (sctp->src_port << 16) | sctp->dest_port; + sctp->checksum = htonl(0); + if (protlen >= + (int)(sizeof(struct sctphdr) + sizeof(struct sctp_chunkhdr))) { + chk = (struct sctp_chunkhdr *)(sctp + 1); + chk->chunk_type = SCTP_SHUTDOWN_ACK; + chk->chunk_flags = 0; + chk->chunk_length = htons(4); + } + if (protlen >= + (int)(sizeof(struct sctphdr) + 2 * sizeof(struct sctp_chunkhdr))) { + chk = chk + 1; + chk->chunk_type = SCTP_PAD_CHUNK; + chk->chunk_flags = 0; + chk->chunk_length = htons(protlen - + (sizeof(struct sctphdr) + sizeof(struct sctp_chunkhdr))); + } + if (doipcksum) { + sctp->checksum = sctp_crc32c(sctp, protlen); + } +} + +int +sctp_check(const u_char *data, int seq) +{ + struct sctphdr *const sctp = (struct sctphdr *) data; + + return (ntohs(sctp->src_port) == ident + && ntohs(sctp->dest_port) == port + (fixedPort ? 0 : seq) + && sctp->v_tag == + (u_int32_t)((sctp->src_port << 16) | sctp->dest_port)); +} + +void gre_prep(struct outdata *outdata) { struct grehdr *const gre = (struct grehdr *) outp; gre->flags = htons(0x2001); gre->proto = htons(port); gre->length = 0; gre->callId = htons(ident + outdata->seq); } int gre_check(const u_char *data, int seq) { struct grehdr *const gre = (struct grehdr *) data; return(ntohs(gre->proto) == port && ntohs(gre->callId) == ident + seq); } void gen_prep(struct outdata *outdata) { u_int16_t *const ptr = (u_int16_t *) outp; ptr[0] = htons(ident); ptr[1] = htons(port + outdata->seq); } int gen_check(const u_char *data, int seq) { u_int16_t *const ptr = (u_int16_t *) data; return(ntohs(ptr[0]) == ident && ntohs(ptr[1]) == port + seq); } void print(register u_char *buf, register int cc, register struct sockaddr_in *from) { register struct ip *ip; register int hlen; char addr[INET_ADDRSTRLEN]; ip = (struct ip *) buf; hlen = ip->ip_hl << 2; cc -= hlen; strlcpy(addr, inet_ntoa(from->sin_addr), sizeof(addr)); if (as_path) Printf(" [AS%u]", as_lookup(asn, addr, AF_INET)); if (nflag) Printf(" %s", addr); else Printf(" %s (%s)", inetname(from->sin_addr), addr); if (verbose) Printf(" %d bytes to %s", cc, inet_ntoa (ip->ip_dst)); } /* * Checksum routine for UDP and TCP headers. */ -u_short -p_cksum(struct ip *ip, u_short *data, int len) +u_short +p_cksum(struct ip *ip, u_short *data, int len, int cov) { static struct ipovly ipo; u_short sum[2]; ipo.ih_pr = ip->ip_p; ipo.ih_len = htons(len); ipo.ih_src = ip->ip_src; ipo.ih_dst = ip->ip_dst; sum[1] = in_cksum((u_short*)&ipo, sizeof(ipo)); /* pseudo ip hdr cksum */ - sum[0] = in_cksum(data, len); /* payload data cksum */ + sum[0] = in_cksum(data, cov); /* payload data cksum */ return ~in_cksum(sum, sizeof(sum)); } /* * Checksum routine for Internet Protocol family headers (C Version) */ u_short in_cksum(register u_short *addr, register int len) { register int nleft = len; register u_short *w = addr; register u_short answer; register int sum = 0; /* * Our algorithm is simple, using a 32 bit accumulator (sum), * we add sequential 16 bit words to it, and at the end, fold * back all the carry bits from the top 16 bits into the lower * 16 bits. */ while (nleft > 1) { sum += *w++; nleft -= 2; } /* mop up an odd byte, if necessary */ if (nleft == 1) sum += *(u_char *)w; /* * add back carry outs from top 16 bits to low 16 bits */ sum = (sum >> 16) + (sum & 0xffff); /* add hi 16 to low 16 */ sum += (sum >> 16); /* add carry */ answer = ~sum; /* truncate to 16 bits */ return (answer); +} + +/* + * CRC32C routine for the Stream Control Transmission Protocol + */ + +#define CRC32C(c, d) (c = (c>>8) ^ crc_c[(c^(d))&0xFF]) + +static u_int32_t crc_c[256] = { + 0x00000000, 0xF26B8303, 0xE13B70F7, 0x1350F3F4, + 0xC79A971F, 0x35F1141C, 0x26A1E7E8, 0xD4CA64EB, + 0x8AD958CF, 0x78B2DBCC, 0x6BE22838, 0x9989AB3B, + 0x4D43CFD0, 0xBF284CD3, 0xAC78BF27, 0x5E133C24, + 0x105EC76F, 0xE235446C, 0xF165B798, 0x030E349B, + 0xD7C45070, 0x25AFD373, 0x36FF2087, 0xC494A384, + 0x9A879FA0, 0x68EC1CA3, 0x7BBCEF57, 0x89D76C54, + 0x5D1D08BF, 0xAF768BBC, 0xBC267848, 0x4E4DFB4B, + 0x20BD8EDE, 0xD2D60DDD, 0xC186FE29, 0x33ED7D2A, + 0xE72719C1, 0x154C9AC2, 0x061C6936, 0xF477EA35, + 0xAA64D611, 0x580F5512, 0x4B5FA6E6, 0xB93425E5, + 0x6DFE410E, 0x9F95C20D, 0x8CC531F9, 0x7EAEB2FA, + 0x30E349B1, 0xC288CAB2, 0xD1D83946, 0x23B3BA45, + 0xF779DEAE, 0x05125DAD, 0x1642AE59, 0xE4292D5A, + 0xBA3A117E, 0x4851927D, 0x5B016189, 0xA96AE28A, + 0x7DA08661, 0x8FCB0562, 0x9C9BF696, 0x6EF07595, + 0x417B1DBC, 0xB3109EBF, 0xA0406D4B, 0x522BEE48, + 0x86E18AA3, 0x748A09A0, 0x67DAFA54, 0x95B17957, + 0xCBA24573, 0x39C9C670, 0x2A993584, 0xD8F2B687, + 0x0C38D26C, 0xFE53516F, 0xED03A29B, 0x1F682198, + 0x5125DAD3, 0xA34E59D0, 0xB01EAA24, 0x42752927, + 0x96BF4DCC, 0x64D4CECF, 0x77843D3B, 0x85EFBE38, + 0xDBFC821C, 0x2997011F, 0x3AC7F2EB, 0xC8AC71E8, + 0x1C661503, 0xEE0D9600, 0xFD5D65F4, 0x0F36E6F7, + 0x61C69362, 0x93AD1061, 0x80FDE395, 0x72966096, + 0xA65C047D, 0x5437877E, 0x4767748A, 0xB50CF789, + 0xEB1FCBAD, 0x197448AE, 0x0A24BB5A, 0xF84F3859, + 0x2C855CB2, 0xDEEEDFB1, 0xCDBE2C45, 0x3FD5AF46, + 0x7198540D, 0x83F3D70E, 0x90A324FA, 0x62C8A7F9, + 0xB602C312, 0x44694011, 0x5739B3E5, 0xA55230E6, + 0xFB410CC2, 0x092A8FC1, 0x1A7A7C35, 0xE811FF36, + 0x3CDB9BDD, 0xCEB018DE, 0xDDE0EB2A, 0x2F8B6829, + 0x82F63B78, 0x709DB87B, 0x63CD4B8F, 0x91A6C88C, + 0x456CAC67, 0xB7072F64, 0xA457DC90, 0x563C5F93, + 0x082F63B7, 0xFA44E0B4, 0xE9141340, 0x1B7F9043, + 0xCFB5F4A8, 0x3DDE77AB, 0x2E8E845F, 0xDCE5075C, + 0x92A8FC17, 0x60C37F14, 0x73938CE0, 0x81F80FE3, + 0x55326B08, 0xA759E80B, 0xB4091BFF, 0x466298FC, + 0x1871A4D8, 0xEA1A27DB, 0xF94AD42F, 0x0B21572C, + 0xDFEB33C7, 0x2D80B0C4, 0x3ED04330, 0xCCBBC033, + 0xA24BB5A6, 0x502036A5, 0x4370C551, 0xB11B4652, + 0x65D122B9, 0x97BAA1BA, 0x84EA524E, 0x7681D14D, + 0x2892ED69, 0xDAF96E6A, 0xC9A99D9E, 0x3BC21E9D, + 0xEF087A76, 0x1D63F975, 0x0E330A81, 0xFC588982, + 0xB21572C9, 0x407EF1CA, 0x532E023E, 0xA145813D, + 0x758FE5D6, 0x87E466D5, 0x94B49521, 0x66DF1622, + 0x38CC2A06, 0xCAA7A905, 0xD9F75AF1, 0x2B9CD9F2, + 0xFF56BD19, 0x0D3D3E1A, 0x1E6DCDEE, 0xEC064EED, + 0xC38D26C4, 0x31E6A5C7, 0x22B65633, 0xD0DDD530, + 0x0417B1DB, 0xF67C32D8, 0xE52CC12C, 0x1747422F, + 0x49547E0B, 0xBB3FFD08, 0xA86F0EFC, 0x5A048DFF, + 0x8ECEE914, 0x7CA56A17, 0x6FF599E3, 0x9D9E1AE0, + 0xD3D3E1AB, 0x21B862A8, 0x32E8915C, 0xC083125F, + 0x144976B4, 0xE622F5B7, 0xF5720643, 0x07198540, + 0x590AB964, 0xAB613A67, 0xB831C993, 0x4A5A4A90, + 0x9E902E7B, 0x6CFBAD78, 0x7FAB5E8C, 0x8DC0DD8F, + 0xE330A81A, 0x115B2B19, 0x020BD8ED, 0xF0605BEE, + 0x24AA3F05, 0xD6C1BC06, 0xC5914FF2, 0x37FACCF1, + 0x69E9F0D5, 0x9B8273D6, 0x88D28022, 0x7AB90321, + 0xAE7367CA, 0x5C18E4C9, 0x4F48173D, 0xBD23943E, + 0xF36E6F75, 0x0105EC76, 0x12551F82, 0xE03E9C81, + 0x34F4F86A, 0xC69F7B69, 0xD5CF889D, 0x27A40B9E, + 0x79B737BA, 0x8BDCB4B9, 0x988C474D, 0x6AE7C44E, + 0xBE2DA0A5, 0x4C4623A6, 0x5F16D052, 0xAD7D5351 +}; + +u_int32_t +sctp_crc32c(const void *packet, u_int32_t len) +{ + u_int32_t i, crc32c; + u_int8_t byte0, byte1, byte2, byte3; + const u_int8_t *buf = (const u_int8_t *)packet; + + crc32c = ~0; + for (i = 0; i < len; i++) + CRC32C(crc32c, buf[i]); + crc32c = ~crc32c; + byte0 = crc32c & 0xff; + byte1 = (crc32c>>8) & 0xff; + byte2 = (crc32c>>16) & 0xff; + byte3 = (crc32c>>24) & 0xff; + crc32c = ((byte0 << 24) | (byte1 << 16) | (byte2 << 8) | byte3); + return htonl(crc32c); } /* * Subtract 2 timeval structs: out = out - in. * Out is assumed to be within about LONG_MAX seconds of in. */ void tvsub(register struct timeval *out, register struct timeval *in) { if ((out->tv_usec -= in->tv_usec) < 0) { --out->tv_sec; out->tv_usec += 1000000; } out->tv_sec -= in->tv_sec; } /* * Construct an Internet address representation. * If the nflag has been supplied, give * numeric value, otherwise try for symbolic name. */ char * inetname(struct in_addr in) { register char *cp; register struct hostent *hp; static int first = 1; static char domain[MAXHOSTNAMELEN + 1], line[MAXHOSTNAMELEN + 1]; if (first && !nflag) { first = 0; if (gethostname(domain, sizeof(domain) - 1) < 0) domain[0] = '\0'; else { cp = strchr(domain, '.'); if (cp == NULL) { hp = gethostbyname(domain); if (hp != NULL) cp = strchr(hp->h_name, '.'); } if (cp == NULL) domain[0] = '\0'; else { ++cp; (void)strncpy(domain, cp, sizeof(domain) - 1); domain[sizeof(domain) - 1] = '\0'; } } } if (!nflag && in.s_addr != INADDR_ANY) { hp = gethostbyaddr((char *)&in, sizeof(in), AF_INET); if (hp != NULL) { if ((cp = strchr(hp->h_name, '.')) != NULL && strcmp(cp + 1, domain) == 0) *cp = '\0'; (void)strncpy(line, hp->h_name, sizeof(line) - 1); line[sizeof(line) - 1] = '\0'; return (line); } } return (inet_ntoa(in)); } struct hostinfo * gethostinfo(register char *hostname) { register int n; register struct hostent *hp; register struct hostinfo *hi; register char **p; register u_int32_t addr, *ap; if (strlen(hostname) >= MAXHOSTNAMELEN) { Fprintf(stderr, "%s: hostname \"%.32s...\" is too long\n", prog, hostname); exit(1); } hi = calloc(1, sizeof(*hi)); if (hi == NULL) { Fprintf(stderr, "%s: calloc %s\n", prog, strerror(errno)); exit(1); } addr = inet_addr(hostname); if ((int32_t)addr != -1) { hi->name = strdup(hostname); hi->n = 1; hi->addrs = calloc(1, sizeof(hi->addrs[0])); if (hi->addrs == NULL) { Fprintf(stderr, "%s: calloc %s\n", prog, strerror(errno)); exit(1); } hi->addrs[0] = addr; return (hi); } hp = gethostbyname(hostname); if (hp == NULL) { Fprintf(stderr, "%s: unknown host %s\n", prog, hostname); exit(1); } if (hp->h_addrtype != AF_INET || hp->h_length != 4) { Fprintf(stderr, "%s: bad host %s\n", prog, hostname); exit(1); } hi->name = strdup(hp->h_name); for (n = 0, p = hp->h_addr_list; *p != NULL; ++n, ++p) continue; hi->n = n; hi->addrs = calloc(n, sizeof(hi->addrs[0])); if (hi->addrs == NULL) { Fprintf(stderr, "%s: calloc %s\n", prog, strerror(errno)); exit(1); } for (ap = hi->addrs, p = hp->h_addr_list; *p != NULL; ++ap, ++p) memcpy(ap, *p, sizeof(*ap)); return (hi); } void freehostinfo(register struct hostinfo *hi) { if (hi->name != NULL) { free(hi->name); hi->name = NULL; } free((char *)hi->addrs); free((char *)hi); } void getaddr(register u_int32_t *ap, register char *hostname) { register struct hostinfo *hi; hi = gethostinfo(hostname); *ap = hi->addrs[0]; freehostinfo(hi); } void setsin(register struct sockaddr_in *sin, register u_int32_t addr) { memset(sin, 0, sizeof(*sin)); #ifdef HAVE_SOCKADDR_SA_LEN sin->sin_len = sizeof(*sin); #endif sin->sin_family = AF_INET; sin->sin_addr.s_addr = addr; } /* String to value with optional min and max. Handles decimal and hex. */ int str2val(register const char *str, register const char *what, register int mi, register int ma) { register const char *cp; register int val; char *ep; if (str[0] == '0' && (str[1] == 'x' || str[1] == 'X')) { cp = str + 2; val = (int)strtol(cp, &ep, 16); } else val = (int)strtol(str, &ep, 10); if (*ep != '\0') { Fprintf(stderr, "%s: \"%s\" bad value for %s \n", prog, str, what); exit(1); } if (val < mi && mi >= 0) { if (mi == 0) Fprintf(stderr, "%s: %s must be >= %d\n", prog, what, mi); else Fprintf(stderr, "%s: %s must be > %d\n", prog, what, mi - 1); exit(1); } if (val > ma && ma >= 0) { Fprintf(stderr, "%s: %s must be <= %d\n", prog, what, ma); exit(1); } return (val); } struct outproto * setproto(char *pname) { struct outproto *proto; int i; for (i = 0; protos[i].name != NULL; i++) { if (strcasecmp(protos[i].name, pname) == 0) { break; } } proto = &protos[i]; if (proto->name == NULL) { /* generic handler */ struct protoent *pe; u_long pnum; /* Determine the IP protocol number */ if ((pe = getprotobyname(pname)) != NULL) pnum = pe->p_proto; else pnum = str2val(optarg, "proto number", 1, 255); proto->num = pnum; } return proto; } void pkt_compare(const u_char *a, int la, const u_char *b, int lb) { int l; int i; for (i = 0; i < la; i++) Printf("%02x", (unsigned int)a[i]); Printf("\n"); l = (la <= lb) ? la : lb; for (i = 0; i < l; i++) if (a[i] == b[i]) Printf("__"); else Printf("%02x", (unsigned int)b[i]); for (; i < lb; i++) Printf("%02x", (unsigned int)b[i]); Printf("\n"); } void usage(void) { extern char version[]; Fprintf(stderr, "Version %s\n", version); Fprintf(stderr, "Usage: %s [-adDeFInrSvx] [-f first_ttl] [-g gateway] [-i iface]\n" "\t[-m max_ttl] [-p port] [-P proto] [-q nqueries] [-s src_addr]\n" "\t[-t tos] [-w waittime] [-A as_server] [-z pausemsecs] host [packetlen]\n", prog); exit(1); } Index: projects/clang-trunk/etc =================================================================== --- projects/clang-trunk/etc (revision 283870) +++ projects/clang-trunk/etc (revision 283871) Property changes on: projects/clang-trunk/etc ___________________________________________________________________ Modified: svn:mergeinfo ## -0,0 +0,1 ## Merged /head/etc:r283596-283870 Index: projects/clang-trunk/lib/libc/arm/SYS.h =================================================================== --- projects/clang-trunk/lib/libc/arm/SYS.h (revision 283870) +++ projects/clang-trunk/lib/libc/arm/SYS.h (revision 283871) @@ -1,84 +1,80 @@ /* $NetBSD: SYS.h,v 1.8 2003/08/07 16:42:02 agc Exp $ */ /*- * Copyright (c) 1990 The Regents of the University of California. * All rights reserved. * * This code is derived from software contributed to Berkeley by * William Jolitz. * * Redistribution and use in source and binary forms, with or without * modification, are permitted provided that the following conditions * are met: * 1. Redistributions of source code must retain the above copyright * notice, this list of conditions and the following disclaimer. * 2. Redistributions in binary form must reproduce the above copyright * notice, this list of conditions and the following disclaimer in the * documentation and/or other materials provided with the distribution. * 3. Neither the name of the University nor the names of its contributors * may be used to endorse or promote products derived from this software * without specific prior written permission. * * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF * SUCH DAMAGE. * * from: @(#)SYS.h 5.5 (Berkeley) 5/7/91 * $FreeBSD$ */ #include #include #include -#ifdef __ARM_EABI__ #define SYSTRAP(x) \ mov ip, r7; \ ldr r7, =SYS_ ## x; \ swi 0; \ mov r7, ip -#else -#define SYSTRAP(x) swi 0 | SYS_ ## x -#endif #define CERROR _C_LABEL(cerror) #define CURBRK _C_LABEL(curbrk) #define _SYSCALL_NOERROR(x) \ ENTRY(__CONCAT(__sys_, x)); \ .weak _C_LABEL(x); \ .set _C_LABEL(x), _C_LABEL(__CONCAT(__sys_,x)); \ .weak _C_LABEL(__CONCAT(_,x)); \ .set _C_LABEL(__CONCAT(_,x)),_C_LABEL(__CONCAT(__sys_,x)); \ SYSTRAP(x) #define _SYSCALL(x) \ _SYSCALL_NOERROR(x); \ it cs; \ bcs PIC_SYM(CERROR, PLT) #define SYSCALL(x) \ _SYSCALL(x) #define PSEUDO(x) \ ENTRY(__CONCAT(__sys_, x)); \ .weak _C_LABEL(__CONCAT(_,x)); \ .set _C_LABEL(__CONCAT(_,x)),_C_LABEL(__CONCAT(__sys_,x)); \ SYSTRAP(x); \ it cs; \ bcs PIC_SYM(CERROR, PLT); \ RET #define RSYSCALL(x) \ _SYSCALL(x); \ RET .globl CERROR Index: projects/clang-trunk/lib/libc/arm/aeabi/aeabi_asm_double.S =================================================================== --- projects/clang-trunk/lib/libc/arm/aeabi/aeabi_asm_double.S (revision 283870) +++ projects/clang-trunk/lib/libc/arm/aeabi/aeabi_asm_double.S (revision 283871) @@ -1,117 +1,119 @@ /* * Copyright (C) 2014 Andrew Turner * All rights reserved. * * Redistribution and use in source and binary forms, with or without * modification, are permitted provided that the following conditions * are met: * 1. Redistributions of source code must retain the above copyright * notice, this list of conditions and the following disclaimer. * 2. Redistributions in binary form must reproduce the above copyright * notice, this list of conditions and the following disclaimer in the * documentation and/or other materials provided with the distribution. * * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE * ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF * SUCH DAMAGE. * */ #include __FBSDID("$FreeBSD$"); #define PCR_Z (1 << 30) #define PCR_C (1 << 29) /* * These functions return the result in the CPSR register. * * For __aeabi_cdcmple: * Z C * LT 0 0 * EQ 1 1 * else 0 1 * * __aeabi_cdrcmple is the same as __aeabi_cdcmple, however the arguments * have been swapped. */ ENTRY(__aeabi_cdcmple) push {r4, r5, r6, r7, ip, lr} /* Backup the input registers */ mov r4, r0 mov r5, r1 mov r6, r2 mov r7, r3 /* Is it less than? */ bl __aeabi_dcmplt cmp r0, #1 bne 1f /* Yes, clear Z and C */ - msr cpsr_c, #(0) + mov ip, #(0) b 99f 1: /* Restore the input regsters for the next function call */ mov r0, r4 mov r1, r5 mov r2, r6 mov r3, r7 /* Is it equal? */ bl __aeabi_dcmpeq cmp r0, #1 bne 2f /* Yes, set Z and C */ - msr cpsr_c, #(PCR_Z | PCR_C) + mov ip, #(PCR_Z | PCR_C) b 99f 2: /* Not less than or equal, set C and clear Z */ - msr cpsr_c, #(PCR_C) + mov ip, #(PCR_C) 99: + msr cpsr_c, ip pop {r4, r5, r6, r7, ip, pc} END(__aeabi_cdcmple) ENTRY(__aeabi_cdrcmple) /* Swap the first half of the arguments */ mov ip, r0 mov r0, r2 mov r2, ip /* And the second half */ mov ip, r1 mov r1, r3 mov r3, ip b __aeabi_cdcmple END(__aeabi_cdrcmple) /* * This is just like __aeabi_cdcmple except it will not throw an exception * in the presence of a quiet NaN. If either argument is a signalling NaN we * will still signal. */ ENTRY(__aeabi_cdcmpeq) /* Check if we can call __aeabi_cfcmple safely */ push {r0, r1, r2, r3, r4, lr} bl __aeabi_cdcmpeq_helper cmp r0, #1 pop {r0, r1, r2, r3, r4, lr} beq 1f bl __aeabi_cdcmple RET 1: - msr cpsr_c, #(PCR_C) + mov ip, #(PCR_C) + msr cpsr_c, ip RET END(__aeabi_cdcmpeq) Index: projects/clang-trunk/lib/libc/arm/aeabi/aeabi_asm_float.S =================================================================== --- projects/clang-trunk/lib/libc/arm/aeabi/aeabi_asm_float.S (revision 283870) +++ projects/clang-trunk/lib/libc/arm/aeabi/aeabi_asm_float.S (revision 283871) @@ -1,108 +1,110 @@ /* * Copyright (C) 2014 Andrew Turner * All rights reserved. * * Redistribution and use in source and binary forms, with or without * modification, are permitted provided that the following conditions * are met: * 1. Redistributions of source code must retain the above copyright * notice, this list of conditions and the following disclaimer. * 2. Redistributions in binary form must reproduce the above copyright * notice, this list of conditions and the following disclaimer in the * documentation and/or other materials provided with the distribution. * * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE * ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF * SUCH DAMAGE. * */ #include __FBSDID("$FreeBSD$"); #define PCR_Z (1 << 30) #define PCR_C (1 << 29) /* * These functions return the result in the CPSR register. * * For __aeabi_cfcmple: * Z C * LT 0 0 * EQ 1 1 * else 0 1 * * __aeabi_cfrcmple is the same as __aeabi_cfcmple, however the arguments * have been swapped. */ ENTRY(__aeabi_cfcmple) push {r4, r5, ip, lr} /* Backup the input registers */ mov r4, r0 mov r5, r1 /* Is it less than? */ bl __aeabi_fcmplt cmp r0, #1 bne 1f /* Yes, clear Z and C */ - msr cpsr_c, #(0) + mov ip, #(0) b 99f 1: /* Restore the input regsters for the next function call */ mov r0, r4 mov r1, r5 /* Is it equal? */ bl __aeabi_fcmpeq cmp r0, #1 bne 2f /* Yes, set Z and C */ - msr cpsr_c, #(PCR_Z | PCR_C) + mov ip, #(PCR_Z | PCR_C) b 99f 2: /* Not less than or equal, set C and clear Z */ - msr cpsr_c, #(PCR_C) + mov ip, #(PCR_C) 99: + msr cpsr_c, ip pop {r4, r5, ip, pc} END(__aeabi_cfcmple) ENTRY(__aeabi_cfrcmple) /* Swap the arguments */ mov ip, r0 mov r0, r1 mov r1, ip b __aeabi_cfcmple END(__aeabi_cfrcmple) /* * This is just like __aeabi_cfcmple except it will not throw an exception * in the presence of a quiet NaN. If either argument is a signalling NaN we * will still signal. */ ENTRY(__aeabi_cfcmpeq) /* Check if we can call __aeabi_cfcmple safely */ push {r0, r1, r2, lr} bl __aeabi_cfcmpeq_helper cmp r0, #1 pop {r0, r1, r2, lr} beq 1f bl __aeabi_cfcmple RET 1: - msreq cpsr_c, #(PCR_C) + mov ip, #(PCR_C) + msr cpsr_c, ip RET END(__aeabi_cfcmpeq) Index: projects/clang-trunk/lib/libc/arm/gen/_setjmp.S =================================================================== --- projects/clang-trunk/lib/libc/arm/gen/_setjmp.S (revision 283870) +++ projects/clang-trunk/lib/libc/arm/gen/_setjmp.S (revision 283871) @@ -1,159 +1,159 @@ /* $NetBSD: _setjmp.S,v 1.12 2013/04/19 13:45:45 matt Exp $ */ /* * Copyright (c) 1997 Mark Brinicombe * All rights reserved. * * Redistribution and use in source and binary forms, with or without * modification, are permitted provided that the following conditions * are met: * 1. Redistributions of source code must retain the above copyright * notice, this list of conditions and the following disclaimer. * 2. Redistributions in binary form must reproduce the above copyright * notice, this list of conditions and the following disclaimer in the * documentation and/or other materials provided with the distribution. * 3. All advertising materials mentioning features or use of this software * must display the following acknowledgement: * This product includes software developed by Mark Brinicombe * 4. Neither the name of the University nor the names of its contributors * may be used to endorse or promote products derived from this software * without specific prior written permission. * * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE * ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF * SUCH DAMAGE. */ #if !defined(__SOFTFP__) && !defined(__VFP_FP__) && !defined(__ARM_PCS) #error FPA is not supported anymore #endif -#if defined(__ARM_EABI__) && !defined(_STANDALONE) +#if !defined(_STANDALONE) .fpu vfp #endif #include #include __FBSDID("$FreeBSD$"); /* * C library -- _setjmp, _longjmp * * _longjmp(a,v) * will generate a "return(v)" from the last call to * _setjmp(a) * by restoring registers from the stack. * The previous signal state is NOT restored. * * Note: r0 is the return value * r1-r3,ip are scratch registers in functions */ ENTRY(_setjmp) ldr r1, .L_setjmp_magic -#if defined(__ARM_EABI__) && !defined(_STANDALONE) +#if !defined(_STANDALONE) ldr r2, .Lfpu_present #ifdef PIC GOT_INIT(r3, .L_setjmp_got, .L_setjmp_gotinit) ldr r2, [r2, r3] #else ldr r2, [r2] #endif teq r2, #0 /* do we have a FPU? */ beq 1f /* no, don't save VFP registers */ orr r1, r1, #(_JB_MAGIC__SETJMP ^ _JB_MAGIC__SETJMP_VFP) /* change magic to VFP magic */ add r2, r0, #(_JB_REG_D8 * 4) vstmia r2, {d8-d15} vmrs r2, fpscr str r2, [r0, #(_JB_REG_FPSCR * 4)] 1: -#endif /* __ARM_EABI__ */ +#endif /* !_STANDALONE */ str r1, [r0] add r0, r0, #(_JB_REG_R4 * 4) /* Store integer registers */ #ifndef __thumb__ stmia r0, {r4-r14} #else stmia r0, {r4-r12} str r13, [r0, #((_JB_REG_R13 - _JB_REG_R4) * 4)] str r14, [r0, #((_JB_REG_R14 - _JB_REG_R4) * 4)] #endif mov r0, #0x00000000 RET END(_setjmp) .L_setjmp_magic: .word _JB_MAGIC__SETJMP -#if defined(__ARM_EABI__) && !defined(_STANDALONE) +#if !defined(_STANDALONE) GOT_INITSYM(.L_setjmp_got, .L_setjmp_gotinit) .Lfpu_present: .word PIC_SYM(_libc_arm_fpu_present, GOTOFF) -#endif /* __ARM_EABI__ */ +#endif /* !_STANDALONE */ WEAK_ALIAS(___longjmp, _longjmp) ENTRY(_longjmp) ldr r2, [r0] /* get magic from jmp_buf */ bic r3, r2, #(_JB_MAGIC__SETJMP ^ _JB_MAGIC__SETJMP_VFP) /* ignore VFP-ness of magic */ ldr ip, .L_setjmp_magic /* load magic */ teq ip, r3 /* magic correct? */ bne botch /* no, botch */ -#if defined(__ARM_EABI__) && !defined(_STANDALONE) +#if !defined(_STANDALONE) teq r3, r2 /* did magic change? */ beq 1f /* no, don't restore VFP */ add ip, r0, #(_JB_REG_D8 * 4) vldmia ip, {d8-d15} ldr ip, [r0, #(_JB_REG_FPSCR * 4)] vmsr fpscr, ip 1: -#endif /* __ARM_EABI__ */ +#endif /* !_STANDALONE */ add r0, r0, #(_JB_REG_R4 * 4) /* Restore integer registers */ #ifndef __thumb__ ldmia r0, {r4-r14} #else ldmia r0, {r4-r12} ldr r13, [r0, #((_JB_REG_R13 - _JB_REG_R4) * 4)] ldr r14, [r0, #((_JB_REG_R14 - _JB_REG_R4) * 4)] #endif /* Validate sp and r14 */ teq sp, #0 it ne teqne r14, #0 it eq beq botch /* Set return value */ movs r0, r1 it eq moveq r0, #0x00000001 RET /* validation failed, die die die. */ botch: #if !defined(_STANDALONE) bl PIC_SYM(_C_LABEL(longjmperror), PLT) bl PIC_SYM(_C_LABEL(abort), PLT) 1: b 1b /* Cannot get here */ #else b . #endif END(_longjmp) Index: projects/clang-trunk/lib/libc/arm/gen/setjmp.S =================================================================== --- projects/clang-trunk/lib/libc/arm/gen/setjmp.S (revision 283870) +++ projects/clang-trunk/lib/libc/arm/gen/setjmp.S (revision 283871) @@ -1,168 +1,160 @@ /* $NetBSD: setjmp.S,v 1.14 2013/04/19 13:45:45 matt Exp $ */ /* * Copyright (c) 1997 Mark Brinicombe * All rights reserved. * * Redistribution and use in source and binary forms, with or without * modification, are permitted provided that the following conditions * are met: * 1. Redistributions of source code must retain the above copyright * notice, this list of conditions and the following disclaimer. * 2. Redistributions in binary form must reproduce the above copyright * notice, this list of conditions and the following disclaimer in the * documentation and/or other materials provided with the distribution. * 3. All advertising materials mentioning features or use of this software * must display the following acknowledgement: * This product includes software developed by Mark Brinicombe * 4. Neither the name of the University nor the names of its contributors * may be used to endorse or promote products derived from this software * without specific prior written permission. * * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE * ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF * SUCH DAMAGE. */ #if !defined(__SOFTFP__) && !defined(__VFP_FP__) && !defined(__ARM_PCS) #error FPA is not supported anymore #endif -#ifdef __ARM_EABI__ .fpu vfp -#endif #include #include __FBSDID("$FreeBSD$"); /* * C library -- setjmp, longjmp * * longjmp(a,v) * will generate a "return(v)" from the last call to * setjmp(a) * by restoring registers from the stack. * The previous signal state is restored. */ ENTRY(setjmp) /* Block all signals and retrieve the old signal mask */ stmfd sp!, {r0, r14} add r2, r0, #(_JB_SIGMASK * 4) /* oset */ mov r1, #0x00000000 /* set */ mov r0, #0x00000001 /* SIG_BLOCK */ bl PIC_SYM(_C_LABEL(sigprocmask), PLT) ldmfd sp!, {r0, r14} ldr r1, .Lsetjmp_magic -#ifdef __ARM_EABI__ ldr r2, .Lfpu_present #ifdef PIC GOT_INIT(r3, .Lsetjmp_got, .Lsetjmp_gotinit) ldr r2, [r2, r3] #else ldr r2, [r2] #endif teq r2, #0 /* do we have a FPU? */ beq 1f /* no, don't save VFP registers */ orr r1, r1, #(_JB_MAGIC_SETJMP ^ _JB_MAGIC_SETJMP_VFP) /* change magic to VFP magic */ add r2, r0, #(_JB_REG_D8 * 4) vstmia r2, {d8-d15} vmrs r2, fpscr str r2, [r0, #(_JB_REG_FPSCR * 4)] 1: -#endif /* __ARM_EABI__ */ str r1, [r0] /* store magic */ /* Store integer registers */ add r0, r0, #(_JB_REG_R4 * 4) #ifndef __thumb__ stmia r0, {r4-r14} #else stmia r0, {r4-r12} str r13, [r0, #((_JB_REG_R13 - _JB_REG_R4) * 4)] str r14, [r0, #((_JB_REG_R14 - _JB_REG_R4) * 4)] #endif mov r0, #0x00000000 RET .Lsetjmp_magic: .word _JB_MAGIC_SETJMP -#ifdef __ARM_EABI__ GOT_INITSYM(.Lsetjmp_got, .Lsetjmp_gotinit) .Lfpu_present: .word PIC_SYM(_libc_arm_fpu_present, GOTOFF) -#endif /* __ARM_EABI__ */ END(setjmp) .weak _C_LABEL(longjmp) .set _C_LABEL(longjmp), _C_LABEL(__longjmp) ENTRY(__longjmp) ldr r2, [r0] ldr ip, .Lsetjmp_magic bic r3, r2, #(_JB_MAGIC_SETJMP ^ _JB_MAGIC_SETJMP_VFP) teq r3, ip bne .Lbotch /* Restore the signal mask. */ stmfd sp!, {r0-r2, r14} mov r2, #0x00000000 add r1, r0, #(_JB_SIGMASK * 4) /* Signal mask */ mov r0, #3 /* SIG_SETMASK */ bl PIC_SYM(_C_LABEL(sigprocmask), PLT) ldmfd sp!, {r0-r2, r14} -#ifdef __ARM_EABI__ tst r2, #(_JB_MAGIC_SETJMP ^ _JB_MAGIC_SETJMP_VFP) /* is this a VFP magic? */ beq 1f /* no, don't restore VFP */ add ip, r0, #(_JB_REG_D8 * 4) vldmia ip, {d8-d15} ldr ip, [r0, #(_JB_REG_FPSCR * 4)] vmsr fpscr, ip 1: -#endif /* __ARM_EABI__ */ add r0, r0, #(_JB_REG_R4 * 4) /* Restore integer registers */ #ifndef __thumb__ ldmia r0, {r4-r14} #else ldmia r0, {r4-r12} ldr r13, [r0, #((_JB_REG_R13 - _JB_REG_R4) * 4)] ldr r14, [r0, #((_JB_REG_R14 - _JB_REG_R4) * 4)] #endif /* Validate sp and r14 */ teq sp, #0 it ne teqne r14, #0 it eq beq .Lbotch /* Set return value */ movs r0, r1 it eq moveq r0, #0x00000001 RET /* validation failed, die die die. */ .Lbotch: bl PIC_SYM(_C_LABEL(longjmperror), PLT) bl PIC_SYM(_C_LABEL(abort), PLT) 1: b 1b /* Cannot get here */ END(__longjmp) Index: projects/clang-trunk/lib/libc/arm/string/ffs.S =================================================================== --- projects/clang-trunk/lib/libc/arm/string/ffs.S (revision 283870) +++ projects/clang-trunk/lib/libc/arm/string/ffs.S (revision 283871) @@ -1,85 +1,86 @@ /* $NetBSD: ffs.S,v 1.5 2003/04/05 23:08:52 bjh21 Exp $ */ /* * Copyright (c) 2001 Christopher Gilbert * All rights reserved. * * Redistribution and use in source and binary forms, with or without * modification, are permitted provided that the following conditions * are met: * 1. Redistributions of source code must retain the above copyright * notice, this list of conditions and the following disclaimer. * 2. Redistributions in binary form must reproduce the above copyright * notice, this list of conditions and the following disclaimer in the * documentation and/or other materials provided with the distribution. * 3. The name of the company nor the name of the author may be used to * endorse or promote products derived from this software without specific * prior written permission. * * THIS SOFTWARE IS PROVIDED BY THE AUTHOR ``AS IS'' AND ANY EXPRESS OR * IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES * OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. * IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, * INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES * (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR * SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF * SUCH DAMAGE. */ #include __FBSDID("$FreeBSD$"); .syntax unified /* * ffs - find first set bit, this algorithm isolates the first set * bit, then multiplies the number by 0x0450fbaf which leaves the top * 6 bits as an index into the table. This algorithm should be a win * over the checking each bit in turn as per the C compiled version. * * under ARMv5 there's an instruction called CLZ (count leading Zero's) that * could be used * * This is the ffs algorithm devised by d.seal and posted to comp.sys.arm on * 16 Feb 1994. */ ENTRY(ffs) /* Standard trick to isolate bottom bit in r0 or 0 if r0 = 0 on entry */ rsb r1, r0, #0 ands r0, r0, r1 #ifndef _ARM_ARCH_5 /* * now r0 has at most one set bit, call this X * if X = 0, all further instructions are skipped */ adrne r2, .L_ffs_table orrne r0, r0, r0, lsl #4 /* r0 = X * 0x11 */ orrne r0, r0, r0, lsl #6 /* r0 = X * 0x451 */ rsbne r0, r0, r0, lsl #16 /* r0 = X * 0x0450fbaf */ /* now lookup in table indexed on top 6 bits of r0 */ ldrbne r0, [ r2, r0, lsr #26 ] RET .text; .type .L_ffs_table, _ASM_TYPE_OBJECT; .L_ffs_table: /* 0 1 2 3 4 5 6 7 */ .byte 0, 1, 2, 13, 3, 7, 0, 14 /* 0- 7 */ .byte 4, 0, 8, 0, 0, 0, 0, 15 /* 8-15 */ .byte 11, 5, 0, 0, 9, 0, 0, 26 /* 16-23 */ .byte 0, 0, 0, 0, 0, 22, 28, 16 /* 24-31 */ .byte 32, 12, 6, 0, 0, 0, 0, 0 /* 32-39 */ .byte 10, 0, 0, 25, 0, 0, 21, 27 /* 40-47 */ .byte 31, 0, 0, 0, 0, 24, 0, 20 /* 48-55 */ .byte 30, 0, 23, 19, 29, 18, 17, 0 /* 56-63 */ #else + itt ne clzne r0, r0 rsbne r0, r0, #32 RET #endif END(ffs) Index: projects/clang-trunk/lib/libc/arm/string/memmove.S =================================================================== --- projects/clang-trunk/lib/libc/arm/string/memmove.S (revision 283870) +++ projects/clang-trunk/lib/libc/arm/string/memmove.S (revision 283871) @@ -1,589 +1,611 @@ /* $NetBSD: memmove.S,v 1.4 2003/10/14 07:51:45 scw Exp $ */ /*- * Copyright (c) 1997 The NetBSD Foundation, Inc. * All rights reserved. * * This code is derived from software contributed to The NetBSD Foundation * by Neil A. Carson and Mark Brinicombe * * Redistribution and use in source and binary forms, with or without * modification, are permitted provided that the following conditions * are met: * 1. Redistributions of source code must retain the above copyright * notice, this list of conditions and the following disclaimer. * 2. Redistributions in binary form must reproduce the above copyright * notice, this list of conditions and the following disclaimer in the * documentation and/or other materials provided with the distribution. * * THIS SOFTWARE IS PROVIDED BY THE NETBSD FOUNDATION, INC. AND CONTRIBUTORS * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR * PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE FOUNDATION OR CONTRIBUTORS * BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE * POSSIBILITY OF SUCH DAMAGE. */ #include __FBSDID("$FreeBSD$"); .syntax unified #ifndef _BCOPY /* LINTSTUB: Func: void *memmove(void *, const void *, size_t) */ ENTRY(memmove) #else /* bcopy = memcpy/memmove with arguments reversed. */ /* LINTSTUB: Func: void bcopy(void *, void *, size_t) */ ENTRY(bcopy) /* switch the source and destination registers */ eor r0, r1, r0 eor r1, r0, r1 eor r0, r1, r0 #endif /* Do the buffers overlap? */ cmp r0, r1 + it eq RETeq /* Bail now if src/dst are the same */ + ite cc subcc r3, r0, r1 /* if (dst > src) r3 = dst - src */ subcs r3, r1, r0 /* if (src > dsr) r3 = src - dst */ cmp r3, r2 /* if (r3 < len) we have an overlap */ bcc PIC_SYM(_C_LABEL(memcpy), PLT) /* Determine copy direction */ cmp r1, r0 + it cc bcc .Lmemmove_backwards + itt eq moveq r0, #0 /* Quick abort for len=0 */ RETeq stmdb sp!, {r0, lr} /* memmove() returns dest addr */ subs r2, r2, #4 blt .Lmemmove_fl4 /* less than 4 bytes */ ands r12, r0, #3 bne .Lmemmove_fdestul /* oh unaligned destination addr */ ands r12, r1, #3 bne .Lmemmove_fsrcul /* oh unaligned source addr */ .Lmemmove_ft8: /* We have aligned source and destination */ subs r2, r2, #8 blt .Lmemmove_fl12 /* less than 12 bytes (4 from above) */ subs r2, r2, #0x14 blt .Lmemmove_fl32 /* less than 32 bytes (12 from above) */ stmdb sp!, {r4} /* borrow r4 */ /* blat 32 bytes at a time */ /* XXX for really big copies perhaps we should use more registers */ .Lmemmove_floop32: ldmia r1!, {r3, r4, r12, lr} stmia r0!, {r3, r4, r12, lr} ldmia r1!, {r3, r4, r12, lr} stmia r0!, {r3, r4, r12, lr} subs r2, r2, #0x20 bge .Lmemmove_floop32 cmn r2, #0x10 + ittt ge ldmiage r1!, {r3, r4, r12, lr} /* blat a remaining 16 bytes */ stmiage r0!, {r3, r4, r12, lr} subge r2, r2, #0x10 ldmia sp!, {r4} /* return r4 */ .Lmemmove_fl32: adds r2, r2, #0x14 /* blat 12 bytes at a time */ .Lmemmove_floop12: + ittt ge ldmiage r1!, {r3, r12, lr} stmiage r0!, {r3, r12, lr} subsge r2, r2, #0x0c bge .Lmemmove_floop12 .Lmemmove_fl12: adds r2, r2, #8 blt .Lmemmove_fl4 subs r2, r2, #4 + itt lt ldrlt r3, [r1], #4 strlt r3, [r0], #4 + ittt ge ldmiage r1!, {r3, r12} stmiage r0!, {r3, r12} subge r2, r2, #4 .Lmemmove_fl4: /* less than 4 bytes to go */ adds r2, r2, #4 + it eq ldmiaeq sp!, {r0, pc} /* done */ /* copy the crud byte at a time */ cmp r2, #2 ldrb r3, [r1], #1 strb r3, [r0], #1 + itt ge ldrbge r3, [r1], #1 strbge r3, [r0], #1 + itt gt ldrbgt r3, [r1], #1 strbgt r3, [r0], #1 ldmia sp!, {r0, pc} /* erg - unaligned destination */ .Lmemmove_fdestul: rsb r12, r12, #4 cmp r12, #2 /* align destination with byte copies */ ldrb r3, [r1], #1 strb r3, [r0], #1 + itt ge ldrbge r3, [r1], #1 strbge r3, [r0], #1 + itt gt ldrbgt r3, [r1], #1 strbgt r3, [r0], #1 subs r2, r2, r12 blt .Lmemmove_fl4 /* less the 4 bytes */ ands r12, r1, #3 beq .Lmemmove_ft8 /* we have an aligned source */ /* erg - unaligned source */ /* This is where it gets nasty ... */ .Lmemmove_fsrcul: bic r1, r1, #3 ldr lr, [r1], #4 cmp r12, #2 bgt .Lmemmove_fsrcul3 beq .Lmemmove_fsrcul2 cmp r2, #0x0c blt .Lmemmove_fsrcul1loop4 sub r2, r2, #0x0c stmdb sp!, {r4, r5} .Lmemmove_fsrcul1loop16: #ifdef __ARMEB__ mov r3, lr, lsl #8 #else mov r3, lr, lsr #8 #endif ldmia r1!, {r4, r5, r12, lr} #ifdef __ARMEB__ orr r3, r3, r4, lsr #24 mov r4, r4, lsl #8 orr r4, r4, r5, lsr #24 mov r5, r5, lsl #8 orr r5, r5, r12, lsr #24 mov r12, r12, lsl #8 orr r12, r12, lr, lsr #24 #else orr r3, r3, r4, lsl #24 mov r4, r4, lsr #8 orr r4, r4, r5, lsl #24 mov r5, r5, lsr #8 orr r5, r5, r12, lsl #24 mov r12, r12, lsr #8 orr r12, r12, lr, lsl #24 #endif stmia r0!, {r3-r5, r12} subs r2, r2, #0x10 bge .Lmemmove_fsrcul1loop16 ldmia sp!, {r4, r5} adds r2, r2, #0x0c blt .Lmemmove_fsrcul1l4 .Lmemmove_fsrcul1loop4: #ifdef __ARMEB__ mov r12, lr, lsl #8 #else mov r12, lr, lsr #8 #endif ldr lr, [r1], #4 #ifdef __ARMEB__ orr r12, r12, lr, lsr #24 #else orr r12, r12, lr, lsl #24 #endif str r12, [r0], #4 subs r2, r2, #4 bge .Lmemmove_fsrcul1loop4 .Lmemmove_fsrcul1l4: sub r1, r1, #3 b .Lmemmove_fl4 .Lmemmove_fsrcul2: cmp r2, #0x0c blt .Lmemmove_fsrcul2loop4 sub r2, r2, #0x0c stmdb sp!, {r4, r5} .Lmemmove_fsrcul2loop16: #ifdef __ARMEB__ mov r3, lr, lsl #16 #else mov r3, lr, lsr #16 #endif ldmia r1!, {r4, r5, r12, lr} #ifdef __ARMEB__ orr r3, r3, r4, lsr #16 mov r4, r4, lsl #16 orr r4, r4, r5, lsr #16 mov r5, r5, lsl #16 orr r5, r5, r12, lsr #16 mov r12, r12, lsl #16 orr r12, r12, lr, lsr #16 #else orr r3, r3, r4, lsl #16 mov r4, r4, lsr #16 orr r4, r4, r5, lsl #16 mov r5, r5, lsr #16 orr r5, r5, r12, lsl #16 mov r12, r12, lsr #16 orr r12, r12, lr, lsl #16 #endif stmia r0!, {r3-r5, r12} subs r2, r2, #0x10 bge .Lmemmove_fsrcul2loop16 ldmia sp!, {r4, r5} adds r2, r2, #0x0c blt .Lmemmove_fsrcul2l4 .Lmemmove_fsrcul2loop4: #ifdef __ARMEB__ mov r12, lr, lsl #16 #else mov r12, lr, lsr #16 #endif ldr lr, [r1], #4 #ifdef __ARMEB__ orr r12, r12, lr, lsr #16 #else orr r12, r12, lr, lsl #16 #endif str r12, [r0], #4 subs r2, r2, #4 bge .Lmemmove_fsrcul2loop4 .Lmemmove_fsrcul2l4: sub r1, r1, #2 b .Lmemmove_fl4 .Lmemmove_fsrcul3: cmp r2, #0x0c blt .Lmemmove_fsrcul3loop4 sub r2, r2, #0x0c stmdb sp!, {r4, r5} .Lmemmove_fsrcul3loop16: #ifdef __ARMEB__ mov r3, lr, lsl #24 #else mov r3, lr, lsr #24 #endif ldmia r1!, {r4, r5, r12, lr} #ifdef __ARMEB__ orr r3, r3, r4, lsr #8 mov r4, r4, lsl #24 orr r4, r4, r5, lsr #8 mov r5, r5, lsl #24 orr r5, r5, r12, lsr #8 mov r12, r12, lsl #24 orr r12, r12, lr, lsr #8 #else orr r3, r3, r4, lsl #8 mov r4, r4, lsr #24 orr r4, r4, r5, lsl #8 mov r5, r5, lsr #24 orr r5, r5, r12, lsl #8 mov r12, r12, lsr #24 orr r12, r12, lr, lsl #8 #endif stmia r0!, {r3-r5, r12} subs r2, r2, #0x10 bge .Lmemmove_fsrcul3loop16 ldmia sp!, {r4, r5} adds r2, r2, #0x0c blt .Lmemmove_fsrcul3l4 .Lmemmove_fsrcul3loop4: #ifdef __ARMEB__ mov r12, lr, lsl #24 #else mov r12, lr, lsr #24 #endif ldr lr, [r1], #4 #ifdef __ARMEB__ orr r12, r12, lr, lsr #8 #else orr r12, r12, lr, lsl #8 #endif str r12, [r0], #4 subs r2, r2, #4 bge .Lmemmove_fsrcul3loop4 .Lmemmove_fsrcul3l4: sub r1, r1, #1 b .Lmemmove_fl4 .Lmemmove_backwards: add r1, r1, r2 add r0, r0, r2 subs r2, r2, #4 blt .Lmemmove_bl4 /* less than 4 bytes */ ands r12, r0, #3 bne .Lmemmove_bdestul /* oh unaligned destination addr */ ands r12, r1, #3 bne .Lmemmove_bsrcul /* oh unaligned source addr */ .Lmemmove_bt8: /* We have aligned source and destination */ subs r2, r2, #8 blt .Lmemmove_bl12 /* less than 12 bytes (4 from above) */ stmdb sp!, {r4, lr} subs r2, r2, #0x14 /* less than 32 bytes (12 from above) */ blt .Lmemmove_bl32 /* blat 32 bytes at a time */ /* XXX for really big copies perhaps we should use more registers */ .Lmemmove_bloop32: ldmdb r1!, {r3, r4, r12, lr} stmdb r0!, {r3, r4, r12, lr} ldmdb r1!, {r3, r4, r12, lr} stmdb r0!, {r3, r4, r12, lr} subs r2, r2, #0x20 bge .Lmemmove_bloop32 .Lmemmove_bl32: cmn r2, #0x10 + ittt ge ldmdbge r1!, {r3, r4, r12, lr} /* blat a remaining 16 bytes */ stmdbge r0!, {r3, r4, r12, lr} subge r2, r2, #0x10 adds r2, r2, #0x14 + ittt ge ldmdbge r1!, {r3, r12, lr} /* blat a remaining 12 bytes */ stmdbge r0!, {r3, r12, lr} subge r2, r2, #0x0c ldmia sp!, {r4, lr} .Lmemmove_bl12: adds r2, r2, #8 blt .Lmemmove_bl4 subs r2, r2, #4 + itt lt ldrlt r3, [r1, #-4]! strlt r3, [r0, #-4]! + ittt ge ldmdbge r1!, {r3, r12} stmdbge r0!, {r3, r12} subge r2, r2, #4 .Lmemmove_bl4: /* less than 4 bytes to go */ adds r2, r2, #4 + it eq RETeq /* done */ /* copy the crud byte at a time */ cmp r2, #2 ldrb r3, [r1, #-1]! strb r3, [r0, #-1]! + itt ge ldrbge r3, [r1, #-1]! strbge r3, [r0, #-1]! + itt gt ldrbgt r3, [r1, #-1]! strbgt r3, [r0, #-1]! RET /* erg - unaligned destination */ .Lmemmove_bdestul: cmp r12, #2 /* align destination with byte copies */ ldrb r3, [r1, #-1]! strb r3, [r0, #-1]! + itt ge ldrbge r3, [r1, #-1]! strbge r3, [r0, #-1]! + itt gt ldrbgt r3, [r1, #-1]! strbgt r3, [r0, #-1]! subs r2, r2, r12 blt .Lmemmove_bl4 /* less than 4 bytes to go */ ands r12, r1, #3 beq .Lmemmove_bt8 /* we have an aligned source */ /* erg - unaligned source */ /* This is where it gets nasty ... */ .Lmemmove_bsrcul: bic r1, r1, #3 ldr r3, [r1, #0] cmp r12, #2 blt .Lmemmove_bsrcul1 beq .Lmemmove_bsrcul2 cmp r2, #0x0c blt .Lmemmove_bsrcul3loop4 sub r2, r2, #0x0c stmdb sp!, {r4, r5, lr} .Lmemmove_bsrcul3loop16: #ifdef __ARMEB__ mov lr, r3, lsr #8 #else mov lr, r3, lsl #8 #endif ldmdb r1!, {r3-r5, r12} #ifdef __ARMEB__ orr lr, lr, r12, lsl #24 mov r12, r12, lsr #8 orr r12, r12, r5, lsl #24 mov r5, r5, lsr #8 orr r5, r5, r4, lsl #24 mov r4, r4, lsr #8 orr r4, r4, r3, lsl #24 #else orr lr, lr, r12, lsr #24 mov r12, r12, lsl #8 orr r12, r12, r5, lsr #24 mov r5, r5, lsl #8 orr r5, r5, r4, lsr #24 mov r4, r4, lsl #8 orr r4, r4, r3, lsr #24 #endif stmdb r0!, {r4, r5, r12, lr} subs r2, r2, #0x10 bge .Lmemmove_bsrcul3loop16 ldmia sp!, {r4, r5, lr} adds r2, r2, #0x0c blt .Lmemmove_bsrcul3l4 .Lmemmove_bsrcul3loop4: #ifdef __ARMEB__ mov r12, r3, lsr #8 #else mov r12, r3, lsl #8 #endif ldr r3, [r1, #-4]! #ifdef __ARMEB__ orr r12, r12, r3, lsl #24 #else orr r12, r12, r3, lsr #24 #endif str r12, [r0, #-4]! subs r2, r2, #4 bge .Lmemmove_bsrcul3loop4 .Lmemmove_bsrcul3l4: add r1, r1, #3 b .Lmemmove_bl4 .Lmemmove_bsrcul2: cmp r2, #0x0c blt .Lmemmove_bsrcul2loop4 sub r2, r2, #0x0c stmdb sp!, {r4, r5, lr} .Lmemmove_bsrcul2loop16: #ifdef __ARMEB__ mov lr, r3, lsr #16 #else mov lr, r3, lsl #16 #endif ldmdb r1!, {r3-r5, r12} #ifdef __ARMEB__ orr lr, lr, r12, lsl #16 mov r12, r12, lsr #16 orr r12, r12, r5, lsl #16 mov r5, r5, lsr #16 orr r5, r5, r4, lsl #16 mov r4, r4, lsr #16 orr r4, r4, r3, lsl #16 #else orr lr, lr, r12, lsr #16 mov r12, r12, lsl #16 orr r12, r12, r5, lsr #16 mov r5, r5, lsl #16 orr r5, r5, r4, lsr #16 mov r4, r4, lsl #16 orr r4, r4, r3, lsr #16 #endif stmdb r0!, {r4, r5, r12, lr} subs r2, r2, #0x10 bge .Lmemmove_bsrcul2loop16 ldmia sp!, {r4, r5, lr} adds r2, r2, #0x0c blt .Lmemmove_bsrcul2l4 .Lmemmove_bsrcul2loop4: #ifdef __ARMEB__ mov r12, r3, lsr #16 #else mov r12, r3, lsl #16 #endif ldr r3, [r1, #-4]! #ifdef __ARMEB__ orr r12, r12, r3, lsl #16 #else orr r12, r12, r3, lsr #16 #endif str r12, [r0, #-4]! subs r2, r2, #4 bge .Lmemmove_bsrcul2loop4 .Lmemmove_bsrcul2l4: add r1, r1, #2 b .Lmemmove_bl4 .Lmemmove_bsrcul1: cmp r2, #0x0c blt .Lmemmove_bsrcul1loop4 sub r2, r2, #0x0c stmdb sp!, {r4, r5, lr} .Lmemmove_bsrcul1loop32: #ifdef __ARMEB__ mov lr, r3, lsr #24 #else mov lr, r3, lsl #24 #endif ldmdb r1!, {r3-r5, r12} #ifdef __ARMEB__ orr lr, lr, r12, lsl #8 mov r12, r12, lsr #24 orr r12, r12, r5, lsl #8 mov r5, r5, lsr #24 orr r5, r5, r4, lsl #8 mov r4, r4, lsr #24 orr r4, r4, r3, lsl #8 #else orr lr, lr, r12, lsr #8 mov r12, r12, lsl #24 orr r12, r12, r5, lsr #8 mov r5, r5, lsl #24 orr r5, r5, r4, lsr #8 mov r4, r4, lsl #24 orr r4, r4, r3, lsr #8 #endif stmdb r0!, {r4, r5, r12, lr} subs r2, r2, #0x10 bge .Lmemmove_bsrcul1loop32 ldmia sp!, {r4, r5, lr} adds r2, r2, #0x0c blt .Lmemmove_bsrcul1l4 .Lmemmove_bsrcul1loop4: #ifdef __ARMEB__ mov r12, r3, lsr #24 #else mov r12, r3, lsl #24 #endif ldr r3, [r1, #-4]! #ifdef __ARMEB__ orr r12, r12, r3, lsl #8 #else orr r12, r12, r3, lsr #8 #endif str r12, [r0, #-4]! subs r2, r2, #4 bge .Lmemmove_bsrcul1loop4 .Lmemmove_bsrcul1l4: add r1, r1, #1 b .Lmemmove_bl4 #ifndef _BCOPY END(memmove) #else END(bcopy) #endif Index: projects/clang-trunk/lib/libc/arm/string/memset.S =================================================================== --- projects/clang-trunk/lib/libc/arm/string/memset.S (revision 283870) +++ projects/clang-trunk/lib/libc/arm/string/memset.S (revision 283871) @@ -1,243 +1,265 @@ /* $NetBSD: memset.S,v 1.4 2003/10/14 07:51:45 scw Exp $ */ /* * Copyright 2003 Wasabi Systems, Inc. * All rights reserved. * * Written by Steve C. Woodford for Wasabi Systems, Inc. * * Redistribution and use in source and binary forms, with or without * modification, are permitted provided that the following conditions * are met: * 1. Redistributions of source code must retain the above copyright * notice, this list of conditions and the following disclaimer. * 2. Redistributions in binary form must reproduce the above copyright * notice, this list of conditions and the following disclaimer in the * documentation and/or other materials provided with the distribution. * 3. All advertising materials mentioning features or use of this software * must display the following acknowledgement: * This product includes software developed for the NetBSD Project by * Wasabi Systems, Inc. * 4. The name of Wasabi Systems, Inc. may not be used to endorse * or promote products derived from this software without specific prior * written permission. * * THIS SOFTWARE IS PROVIDED BY WASABI SYSTEMS, INC. ``AS IS'' AND * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR * PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL WASABI SYSTEMS, INC * BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE * POSSIBILITY OF SUCH DAMAGE. */ /* * Copyright (c) 1995 Mark Brinicombe. * All rights reserved. * * Redistribution and use in source and binary forms, with or without * modification, are permitted provided that the following conditions * are met: * 1. Redistributions of source code must retain the above copyright * notice, this list of conditions and the following disclaimer. * 2. Redistributions in binary form must reproduce the above copyright * notice, this list of conditions and the following disclaimer in the * documentation and/or other materials provided with the distribution. * 3. All advertising materials mentioning features or use of this software * must display the following acknowledgement: * This product includes software developed by Mark Brinicombe. * 4. The name of the company nor the name of the author may be used to * endorse or promote products derived from this software without specific * prior written permission. * * THIS SOFTWARE IS PROVIDED BY THE AUTHOR ``AS IS'' AND ANY EXPRESS OR IMPLIED * WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF * MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. * IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, * INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES * (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR * SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF * SUCH DAMAGE. */ #include __FBSDID("$FreeBSD$"); .syntax unified /* * memset: Sets a block of memory to the specified value * * On entry: * r0 - dest address * r1 - byte to write * r2 - number of bytes to write * * On exit: * r0 - dest address */ #ifdef _BZERO /* LINTSTUB: Func: void bzero(void *, size_t) */ ENTRY(bzero) mov r3, #0x00 #else /* LINTSTUB: Func: void *memset(void *, int, size_t) */ ENTRY(memset) and r3, r1, #0xff /* We deal with bytes */ mov r1, r2 #endif cmp r1, #0x04 /* Do we have less than 4 bytes */ mov ip, r0 blt .Lmemset_lessthanfour /* Ok first we will word align the address */ ands r2, ip, #0x03 /* Get the bottom two bits */ bne .Lmemset_wordunaligned /* The address is not word aligned */ /* We are now word aligned */ .Lmemset_wordaligned: #ifndef _BZERO orr r3, r3, r3, lsl #8 /* Extend value to 16-bits */ #endif #ifdef _ARM_ARCH_5E tst ip, #0x04 /* Quad-align for armv5e */ #else cmp r1, #0x10 #endif #ifndef _BZERO orr r3, r3, r3, lsl #16 /* Extend value to 32-bits */ #endif #ifdef _ARM_ARCH_5E + itt ne subne r1, r1, #0x04 /* Quad-align if necessary */ strne r3, [ip], #0x04 cmp r1, #0x10 #endif blt .Lmemset_loop4 /* If less than 16 then use words */ mov r2, r3 /* Duplicate data */ cmp r1, #0x80 /* If < 128 then skip the big loop */ blt .Lmemset_loop32 /* Do 128 bytes at a time */ .Lmemset_loop128: subs r1, r1, #0x80 #ifdef _ARM_ARCH_5E + itttt ge strdge r2, [ip], #0x08 strdge r2, [ip], #0x08 strdge r2, [ip], #0x08 strdge r2, [ip], #0x08 + itttt ge strdge r2, [ip], #0x08 strdge r2, [ip], #0x08 strdge r2, [ip], #0x08 strdge r2, [ip], #0x08 + itttt ge strdge r2, [ip], #0x08 strdge r2, [ip], #0x08 strdge r2, [ip], #0x08 strdge r2, [ip], #0x08 + itttt ge strdge r2, [ip], #0x08 strdge r2, [ip], #0x08 strdge r2, [ip], #0x08 strdge r2, [ip], #0x08 #else stmiage ip!, {r2-r3} stmiage ip!, {r2-r3} stmiage ip!, {r2-r3} stmiage ip!, {r2-r3} stmiage ip!, {r2-r3} stmiage ip!, {r2-r3} stmiage ip!, {r2-r3} stmiage ip!, {r2-r3} stmiage ip!, {r2-r3} stmiage ip!, {r2-r3} stmiage ip!, {r2-r3} stmiage ip!, {r2-r3} stmiage ip!, {r2-r3} stmiage ip!, {r2-r3} stmiage ip!, {r2-r3} stmiage ip!, {r2-r3} #endif bgt .Lmemset_loop128 + it eq RETeq /* Zero length so just exit */ add r1, r1, #0x80 /* Adjust for extra sub */ /* Do 32 bytes at a time */ .Lmemset_loop32: subs r1, r1, #0x20 + itttt ge #ifdef _ARM_ARCH_5E strdge r2, [ip], #0x08 strdge r2, [ip], #0x08 strdge r2, [ip], #0x08 strdge r2, [ip], #0x08 #else stmiage ip!, {r2-r3} stmiage ip!, {r2-r3} stmiage ip!, {r2-r3} stmiage ip!, {r2-r3} #endif bgt .Lmemset_loop32 + it eq RETeq /* Zero length so just exit */ adds r1, r1, #0x10 /* Partially adjust for extra sub */ /* Deal with 16 bytes or more */ + itt ge #ifdef _ARM_ARCH_5E strdge r2, [ip], #0x08 strdge r2, [ip], #0x08 #else stmiage ip!, {r2-r3} stmiage ip!, {r2-r3} #endif + it eq RETeq /* Zero length so just exit */ + it lt addlt r1, r1, #0x10 /* Possibly adjust for extra sub */ /* We have at least 4 bytes so copy as words */ .Lmemset_loop4: subs r1, r1, #0x04 + it ge strge r3, [ip], #0x04 bgt .Lmemset_loop4 + it eq RETeq /* Zero length so just exit */ #ifdef _ARM_ARCH_5E /* Compensate for 64-bit alignment check */ adds r1, r1, #0x04 + it eq RETeq cmp r1, #2 #else cmp r1, #-2 #endif strb r3, [ip], #0x01 /* Set 1 byte */ + it ge strbge r3, [ip], #0x01 /* Set another byte */ + it gt strbgt r3, [ip] /* and a third */ RET /* Exit */ .Lmemset_wordunaligned: rsb r2, r2, #0x004 strb r3, [ip], #0x01 /* Set 1 byte */ cmp r2, #0x02 + it ge strbge r3, [ip], #0x01 /* Set another byte */ sub r1, r1, r2 + it gt strbgt r3, [ip], #0x01 /* and a third */ cmp r1, #0x04 /* More than 4 bytes left? */ + it ge bge .Lmemset_wordaligned /* Yup */ .Lmemset_lessthanfour: cmp r1, #0x00 + it eq RETeq /* Zero length so exit */ strb r3, [ip], #0x01 /* Set 1 byte */ cmp r1, #0x02 + it ge strbge r3, [ip], #0x01 /* Set another byte */ + it gt strbgt r3, [ip] /* and a third */ RET /* Exit */ #ifdef _BZERO END(bzero) #else END(memset) #endif Index: projects/clang-trunk/lib/libc/arm/string/strcmp.S =================================================================== --- projects/clang-trunk/lib/libc/arm/string/strcmp.S (revision 283870) +++ projects/clang-trunk/lib/libc/arm/string/strcmp.S (revision 283871) @@ -1,44 +1,45 @@ /* $NetBSD: strcmp.S,v 1.3 2003/04/05 23:08:52 bjh21 Exp $ */ /* * Copyright (c) 2002 ARM Ltd * All rights reserved. * * Redistribution and use in source and binary forms, with or without * modification, are permitted provided that the following conditions * are met: * 1. Redistributions of source code must retain the above copyright * notice, this list of conditions and the following disclaimer. * 2. Redistributions in binary form must reproduce the above copyright * notice, this list of conditions and the following disclaimer in the * documentation and/or other materials provided with the distribution. * 3. The name of the company may not be used to endorse or promote * products derived from this software without specific prior written * permission. * * THIS SOFTWARE IS PROVIDED BY ARM LTD ``AS IS'' AND ANY EXPRESS OR IMPLIED * WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF * MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. * IN NO EVENT SHALL ARM LTD BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, * SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED * TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR * PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF * LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING * NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS * SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. */ #include __FBSDID("$FreeBSD$"); ENTRY(strcmp) 1: ldrb r2, [r0], #1 ldrb r3, [r1], #1 cmp r2, #1 + it cs cmpcs r2, r3 beq 1b sub r0, r2, r3 RET END(strcmp) Index: projects/clang-trunk/lib/libc/arm/string/strlen.S =================================================================== --- projects/clang-trunk/lib/libc/arm/string/strlen.S (revision 283870) +++ projects/clang-trunk/lib/libc/arm/string/strlen.S (revision 283871) @@ -1,81 +1,85 @@ /*- * Copyright (c) 2005 Olivier Houchard * All rights reserved. * * Redistribution and use in source and binary forms, with or without * modification, are permitted provided that the following conditions * are met: * 1. Redistributions of source code must retain the above copyright * notice, this list of conditions and the following disclaimer. * 2. Redistributions in binary form must reproduce the above copyright * notice, this list of conditions and the following disclaimer in the * documentation and/or other materials provided with the distribution. * * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE * ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF * SUCH DAMAGE. */ #include __FBSDID("$FreeBSD$"); .syntax unified ENTRY(strlen) mov r1, #0 /* Check that the pointer is aligned on 32 bits. */ ands r3, r0, #3 beq .Loop sub r0, r0, r3 ldr r2, [r0] add r0, r0, #4 cmp r3, #2 blt .Ldo_3 bgt .Ldo_1 /* So that the N bit is set. */ cmp r3, #0 b .Ldo_2 .Loop: ldr r2, [r0] add r0, r0, #4 #ifndef __ARMEB__ ands r3, r2, #0x000000ff #else ands r3, r2, #0xff000000 #endif + it ne addne r1, r1, #1 .Ldo_3: + itt ne #ifndef __ARMEB__ andsne r3, r2, #0x0000ff00 #else andsne r3, r2, #0x00ff0000 #endif addne r1, r1, #1 .Ldo_2: + itt ne #ifndef __ARMEB__ andsne r3, r2, #0x00ff0000 #else andsne r3, r2, #0x0000ff00 #endif addne r1, r1, #1 .Ldo_1: + ittt ne #ifndef __ARMEB__ andsne r3, r2, #0xff000000 #else andsne r3, r2, #0x000000ff #endif addne r1, r1, #1 bne .Loop .Lexit: mov r0, r1 RET END(strlen) Index: projects/clang-trunk/lib/libc/arm/string/strncmp.S =================================================================== --- projects/clang-trunk/lib/libc/arm/string/strncmp.S (revision 283870) +++ projects/clang-trunk/lib/libc/arm/string/strncmp.S (revision 283871) @@ -1,55 +1,58 @@ /* $NetBSD: strncmp.S,v 1.2 2003/04/05 23:08:52 bjh21 Exp $ */ /* * Copyright (c) 2002 ARM Ltd * All rights reserved. * * Redistribution and use in source and binary forms, with or without * modification, are permitted provided that the following conditions * are met: * 1. Redistributions of source code must retain the above copyright * notice, this list of conditions and the following disclaimer. * 2. Redistributions in binary form must reproduce the above copyright * notice, this list of conditions and the following disclaimer in the * documentation and/or other materials provided with the distribution. * 3. The name of the company may not be used to endorse or promote * products derived from this software without specific prior written * permission. * * THIS SOFTWARE IS PROVIDED BY ARM LTD ``AS IS'' AND ANY EXPRESS OR IMPLIED * WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF * MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. * IN NO EVENT SHALL ARM LTD BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, * SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED * TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR * PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF * LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING * NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS * SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. */ #include __FBSDID("$FreeBSD$"); ENTRY(strncmp) /* if (len == 0) return 0 */ cmp r2, #0 + itt eq moveq r0, #0 moveq pc, lr /* ip == last src address to compare */ adds ip, r0, r2 /* Use last possible address on overflow. */ + it cs movcs ip, #0 sub ip, ip, #1 1: ldrb r2, [r0], #1 ldrb r3, [r1], #1 cmp ip, r0 + itt cs cmpcs r2, #1 cmpcs r2, r3 beq 1b sub r0, r2, r3 RET END(strncmp) Index: projects/clang-trunk/lib/libc/xdr/xdr.c =================================================================== --- projects/clang-trunk/lib/libc/xdr/xdr.c (revision 283870) +++ projects/clang-trunk/lib/libc/xdr/xdr.c (revision 283871) @@ -1,967 +1,907 @@ /* $NetBSD: xdr.c,v 1.22 2000/07/06 03:10:35 christos Exp $ */ /*- * Copyright (c) 2010, Oracle America, Inc. * * Redistribution and use in source and binary forms, with or without * modification, are permitted provided that the following conditions are * met: * * * Redistributions of source code must retain the above copyright * notice, this list of conditions and the following disclaimer. * * Redistributions in binary form must reproduce the above * copyright notice, this list of conditions and the following * disclaimer in the documentation and/or other materials * provided with the distribution. * * Neither the name of the "Oracle America, Inc." nor the names of its * contributors may be used to endorse or promote products derived * from this software without specific prior written permission. * * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS * FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE * COPYRIGHT HOLDER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, * INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE * GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, * WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING * NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE * OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. */ #if defined(LIBC_SCCS) && !defined(lint) static char *sccsid2 = "@(#)xdr.c 1.35 87/08/12"; static char *sccsid = "@(#)xdr.c 2.1 88/07/29 4.0 RPCSRC"; #endif #include __FBSDID("$FreeBSD$"); /* * xdr.c, Generic XDR routines implementation. * * These are the "generic" xdr routines used to serialize and de-serialize * most common data items. See xdr.h for more info on the interface to * xdr. */ #include "namespace.h" #include #include #include #include #include #include #include "un-namespace.h" typedef quad_t longlong_t; /* ANSI long long type */ typedef u_quad_t u_longlong_t; /* ANSI unsigned long long type */ /* * constants specific to the xdr "protocol" */ #define XDR_FALSE ((long) 0) #define XDR_TRUE ((long) 1) #define LASTUNSIGNED ((u_int) 0-1) /* * for unit alignment */ static const char xdr_zero[BYTES_PER_XDR_UNIT] = { 0, 0, 0, 0 }; /* * Free a data structure using XDR * Not a filter, but a convenient utility nonetheless */ void -xdr_free(proc, objp) - xdrproc_t proc; - void *objp; +xdr_free(xdrproc_t proc, void *objp) { XDR x; x.x_op = XDR_FREE; (*proc)(&x, objp); } /* * XDR nothing */ bool_t xdr_void(void) { return (TRUE); } /* * XDR integers */ bool_t -xdr_int(xdrs, ip) - XDR *xdrs; - int *ip; +xdr_int(XDR *xdrs, int *ip) { long l; switch (xdrs->x_op) { case XDR_ENCODE: l = (long) *ip; return (XDR_PUTLONG(xdrs, &l)); case XDR_DECODE: if (!XDR_GETLONG(xdrs, &l)) { return (FALSE); } *ip = (int) l; return (TRUE); case XDR_FREE: return (TRUE); } /* NOTREACHED */ return (FALSE); } /* * XDR unsigned integers */ bool_t -xdr_u_int(xdrs, up) - XDR *xdrs; - u_int *up; +xdr_u_int(XDR *xdrs, u_int *up) { u_long l; switch (xdrs->x_op) { case XDR_ENCODE: l = (u_long) *up; return (XDR_PUTLONG(xdrs, (long *)&l)); case XDR_DECODE: if (!XDR_GETLONG(xdrs, (long *)&l)) { return (FALSE); } *up = (u_int) l; return (TRUE); case XDR_FREE: return (TRUE); } /* NOTREACHED */ return (FALSE); } /* * XDR long integers * same as xdr_u_long - open coded to save a proc call! */ bool_t -xdr_long(xdrs, lp) - XDR *xdrs; - long *lp; +xdr_long(XDR *xdrs, long *lp) { switch (xdrs->x_op) { case XDR_ENCODE: return (XDR_PUTLONG(xdrs, lp)); case XDR_DECODE: return (XDR_GETLONG(xdrs, lp)); case XDR_FREE: return (TRUE); } /* NOTREACHED */ return (FALSE); } /* * XDR unsigned long integers * same as xdr_long - open coded to save a proc call! */ bool_t -xdr_u_long(xdrs, ulp) - XDR *xdrs; - u_long *ulp; +xdr_u_long(XDR *xdrs, u_long *ulp) { switch (xdrs->x_op) { case XDR_ENCODE: return (XDR_PUTLONG(xdrs, (long *)ulp)); case XDR_DECODE: return (XDR_GETLONG(xdrs, (long *)ulp)); case XDR_FREE: return (TRUE); } /* NOTREACHED */ return (FALSE); } /* * XDR 32-bit integers * same as xdr_u_int32_t - open coded to save a proc call! */ bool_t -xdr_int32_t(xdrs, int32_p) - XDR *xdrs; - int32_t *int32_p; +xdr_int32_t(XDR *xdrs, int32_t *int32_p) { long l; switch (xdrs->x_op) { case XDR_ENCODE: l = (long) *int32_p; return (XDR_PUTLONG(xdrs, &l)); case XDR_DECODE: if (!XDR_GETLONG(xdrs, &l)) { return (FALSE); } *int32_p = (int32_t) l; return (TRUE); case XDR_FREE: return (TRUE); } /* NOTREACHED */ return (FALSE); } /* * XDR unsigned 32-bit integers * same as xdr_int32_t - open coded to save a proc call! */ bool_t -xdr_u_int32_t(xdrs, u_int32_p) - XDR *xdrs; - u_int32_t *u_int32_p; +xdr_u_int32_t(XDR *xdrs, u_int32_t *u_int32_p) { u_long l; switch (xdrs->x_op) { case XDR_ENCODE: l = (u_long) *u_int32_p; return (XDR_PUTLONG(xdrs, (long *)&l)); case XDR_DECODE: if (!XDR_GETLONG(xdrs, (long *)&l)) { return (FALSE); } *u_int32_p = (u_int32_t) l; return (TRUE); case XDR_FREE: return (TRUE); } /* NOTREACHED */ return (FALSE); } /* * XDR unsigned 32-bit integers * same as xdr_int32_t - open coded to save a proc call! */ bool_t -xdr_uint32_t(xdrs, u_int32_p) - XDR *xdrs; - uint32_t *u_int32_p; +xdr_uint32_t(XDR *xdrs, uint32_t *u_int32_p) { u_long l; switch (xdrs->x_op) { case XDR_ENCODE: l = (u_long) *u_int32_p; return (XDR_PUTLONG(xdrs, (long *)&l)); case XDR_DECODE: if (!XDR_GETLONG(xdrs, (long *)&l)) { return (FALSE); } *u_int32_p = (u_int32_t) l; return (TRUE); case XDR_FREE: return (TRUE); } /* NOTREACHED */ return (FALSE); } /* * XDR short integers */ bool_t -xdr_short(xdrs, sp) - XDR *xdrs; - short *sp; +xdr_short(XDR *xdrs, short *sp) { long l; switch (xdrs->x_op) { case XDR_ENCODE: l = (long) *sp; return (XDR_PUTLONG(xdrs, &l)); case XDR_DECODE: if (!XDR_GETLONG(xdrs, &l)) { return (FALSE); } *sp = (short) l; return (TRUE); case XDR_FREE: return (TRUE); } /* NOTREACHED */ return (FALSE); } /* * XDR unsigned short integers */ bool_t -xdr_u_short(xdrs, usp) - XDR *xdrs; - u_short *usp; +xdr_u_short(XDR *xdrs, u_short *usp) { u_long l; switch (xdrs->x_op) { case XDR_ENCODE: l = (u_long) *usp; return (XDR_PUTLONG(xdrs, (long *)&l)); case XDR_DECODE: if (!XDR_GETLONG(xdrs, (long *)&l)) { return (FALSE); } *usp = (u_short) l; return (TRUE); case XDR_FREE: return (TRUE); } /* NOTREACHED */ return (FALSE); } /* * XDR 16-bit integers */ bool_t -xdr_int16_t(xdrs, int16_p) - XDR *xdrs; - int16_t *int16_p; +xdr_int16_t(XDR *xdrs, int16_t *int16_p) { long l; switch (xdrs->x_op) { case XDR_ENCODE: l = (long) *int16_p; return (XDR_PUTLONG(xdrs, &l)); case XDR_DECODE: if (!XDR_GETLONG(xdrs, &l)) { return (FALSE); } *int16_p = (int16_t) l; return (TRUE); case XDR_FREE: return (TRUE); } /* NOTREACHED */ return (FALSE); } /* * XDR unsigned 16-bit integers */ bool_t -xdr_u_int16_t(xdrs, u_int16_p) - XDR *xdrs; - u_int16_t *u_int16_p; +xdr_u_int16_t(XDR *xdrs, u_int16_t *u_int16_p) { u_long l; switch (xdrs->x_op) { case XDR_ENCODE: l = (u_long) *u_int16_p; return (XDR_PUTLONG(xdrs, (long *)&l)); case XDR_DECODE: if (!XDR_GETLONG(xdrs, (long *)&l)) { return (FALSE); } *u_int16_p = (u_int16_t) l; return (TRUE); case XDR_FREE: return (TRUE); } /* NOTREACHED */ return (FALSE); } /* * XDR unsigned 16-bit integers */ bool_t -xdr_uint16_t(xdrs, u_int16_p) - XDR *xdrs; - uint16_t *u_int16_p; +xdr_uint16_t(XDR *xdrs, uint16_t *u_int16_p) { u_long l; switch (xdrs->x_op) { case XDR_ENCODE: l = (u_long) *u_int16_p; return (XDR_PUTLONG(xdrs, (long *)&l)); case XDR_DECODE: if (!XDR_GETLONG(xdrs, (long *)&l)) { return (FALSE); } *u_int16_p = (u_int16_t) l; return (TRUE); case XDR_FREE: return (TRUE); } /* NOTREACHED */ return (FALSE); } /* * XDR a char */ bool_t -xdr_char(xdrs, cp) - XDR *xdrs; - char *cp; +xdr_char(XDR *xdrs, char *cp) { int i; i = (*cp); if (!xdr_int(xdrs, &i)) { return (FALSE); } *cp = i; return (TRUE); } /* * XDR an unsigned char */ bool_t -xdr_u_char(xdrs, cp) - XDR *xdrs; - u_char *cp; +xdr_u_char(XDR *xdrs, u_char *cp) { u_int u; u = (*cp); if (!xdr_u_int(xdrs, &u)) { return (FALSE); } *cp = u; return (TRUE); } /* * XDR booleans */ bool_t -xdr_bool(xdrs, bp) - XDR *xdrs; - bool_t *bp; +xdr_bool(XDR *xdrs, bool_t *bp) { long lb; switch (xdrs->x_op) { case XDR_ENCODE: lb = *bp ? XDR_TRUE : XDR_FALSE; return (XDR_PUTLONG(xdrs, &lb)); case XDR_DECODE: if (!XDR_GETLONG(xdrs, &lb)) { return (FALSE); } *bp = (lb == XDR_FALSE) ? FALSE : TRUE; return (TRUE); case XDR_FREE: return (TRUE); } /* NOTREACHED */ return (FALSE); } /* * XDR enumerations */ bool_t -xdr_enum(xdrs, ep) - XDR *xdrs; - enum_t *ep; +xdr_enum(XDR *xdrs, enum_t *ep) { enum sizecheck { SIZEVAL }; /* used to find the size of an enum */ /* * enums are treated as ints */ /* LINTED */ if (sizeof (enum sizecheck) == sizeof (long)) { return (xdr_long(xdrs, (long *)(void *)ep)); } else /* LINTED */ if (sizeof (enum sizecheck) == sizeof (int)) { return (xdr_int(xdrs, (int *)(void *)ep)); } else /* LINTED */ if (sizeof (enum sizecheck) == sizeof (short)) { return (xdr_short(xdrs, (short *)(void *)ep)); } else { return (FALSE); } } /* * XDR opaque data * Allows the specification of a fixed size sequence of opaque bytes. * cp points to the opaque object and cnt gives the byte length. */ bool_t -xdr_opaque(xdrs, cp, cnt) - XDR *xdrs; - caddr_t cp; - u_int cnt; +xdr_opaque(XDR *xdrs, caddr_t cp, u_int cnt) { u_int rndup; static int crud[BYTES_PER_XDR_UNIT]; /* * if no data we are done */ if (cnt == 0) return (TRUE); /* * round byte count to full xdr units */ rndup = cnt % BYTES_PER_XDR_UNIT; if (rndup > 0) rndup = BYTES_PER_XDR_UNIT - rndup; if (xdrs->x_op == XDR_DECODE) { if (!XDR_GETBYTES(xdrs, cp, cnt)) { return (FALSE); } if (rndup == 0) return (TRUE); return (XDR_GETBYTES(xdrs, (caddr_t)(void *)crud, rndup)); } if (xdrs->x_op == XDR_ENCODE) { if (!XDR_PUTBYTES(xdrs, cp, cnt)) { return (FALSE); } if (rndup == 0) return (TRUE); return (XDR_PUTBYTES(xdrs, xdr_zero, rndup)); } if (xdrs->x_op == XDR_FREE) { return (TRUE); } return (FALSE); } /* * XDR counted bytes * *cpp is a pointer to the bytes, *sizep is the count. * If *cpp is NULL maxsize bytes are allocated */ bool_t -xdr_bytes(xdrs, cpp, sizep, maxsize) - XDR *xdrs; - char **cpp; - u_int *sizep; - u_int maxsize; +xdr_bytes(XDR *xdrs, char **cpp, u_int *sizep, u_int maxsize) { char *sp = *cpp; /* sp is the actual string pointer */ u_int nodesize; /* * first deal with the length since xdr bytes are counted */ if (! xdr_u_int(xdrs, sizep)) { return (FALSE); } nodesize = *sizep; if ((nodesize > maxsize) && (xdrs->x_op != XDR_FREE)) { return (FALSE); } /* * now deal with the actual bytes */ switch (xdrs->x_op) { case XDR_DECODE: if (nodesize == 0) { return (TRUE); } if (sp == NULL) { *cpp = sp = mem_alloc(nodesize); } if (sp == NULL) { warnx("xdr_bytes: out of memory"); return (FALSE); } /* FALLTHROUGH */ case XDR_ENCODE: return (xdr_opaque(xdrs, sp, nodesize)); case XDR_FREE: if (sp != NULL) { mem_free(sp, nodesize); *cpp = NULL; } return (TRUE); } /* NOTREACHED */ return (FALSE); } /* * Implemented here due to commonality of the object. */ bool_t -xdr_netobj(xdrs, np) - XDR *xdrs; - struct netobj *np; +xdr_netobj(XDR *xdrs, struct netobj *np) { return (xdr_bytes(xdrs, &np->n_bytes, &np->n_len, MAX_NETOBJ_SZ)); } /* * XDR a descriminated union * Support routine for discriminated unions. * You create an array of xdrdiscrim structures, terminated with * an entry with a null procedure pointer. The routine gets * the discriminant value and then searches the array of xdrdiscrims * looking for that value. It calls the procedure given in the xdrdiscrim * to handle the discriminant. If there is no specific routine a default * routine may be called. * If there is no specific or default routine an error is returned. */ bool_t -xdr_union(xdrs, dscmp, unp, choices, dfault) - XDR *xdrs; - enum_t *dscmp; /* enum to decide which arm to work on */ - char *unp; /* the union itself */ - const struct xdr_discrim *choices; /* [value, xdr proc] for each arm */ - xdrproc_t dfault; /* default xdr routine */ +xdr_union(XDR *xdrs, enum_t *dscmp, char *unp, const struct xdr_discrim *choices, xdrproc_t dfault) +/* + * XDR *xdrs; + * enum_t *dscmp; // enum to decide which arm to work on + * char *unp; // the union itself + * const struct xdr_discrim *choices; // [value, xdr proc] for each arm + * xdrproc_t dfault; // default xdr routine + */ { enum_t dscm; /* * we deal with the discriminator; it's an enum */ if (! xdr_enum(xdrs, dscmp)) { return (FALSE); } dscm = *dscmp; /* * search choices for a value that matches the discriminator. * if we find one, execute the xdr routine for that value. */ for (; choices->proc != NULL_xdrproc_t; choices++) { if (choices->value == dscm) return ((*(choices->proc))(xdrs, unp)); } /* * no match - execute the default xdr routine if there is one */ return ((dfault == NULL_xdrproc_t) ? FALSE : (*dfault)(xdrs, unp)); } /* * Non-portable xdr primitives. * Care should be taken when moving these routines to new architectures. */ /* * XDR null terminated ASCII strings * xdr_string deals with "C strings" - arrays of bytes that are * terminated by a NULL character. The parameter cpp references a * pointer to storage; If the pointer is null, then the necessary * storage is allocated. The last parameter is the max allowed length * of the string as specified by a protocol. */ bool_t -xdr_string(xdrs, cpp, maxsize) - XDR *xdrs; - char **cpp; - u_int maxsize; +xdr_string(XDR *xdrs, char **cpp, u_int maxsize) { char *sp = *cpp; /* sp is the actual string pointer */ u_int size; u_int nodesize; /* * first deal with the length since xdr strings are counted-strings */ switch (xdrs->x_op) { case XDR_FREE: if (sp == NULL) { return(TRUE); /* already free */ } /* FALLTHROUGH */ case XDR_ENCODE: size = strlen(sp); break; case XDR_DECODE: break; } if (! xdr_u_int(xdrs, &size)) { return (FALSE); } if (size > maxsize) { return (FALSE); } nodesize = size + 1; /* * now deal with the actual bytes */ switch (xdrs->x_op) { case XDR_DECODE: if (nodesize == 0) { return (TRUE); } if (sp == NULL) *cpp = sp = mem_alloc(nodesize); if (sp == NULL) { warnx("xdr_string: out of memory"); return (FALSE); } sp[size] = 0; /* FALLTHROUGH */ case XDR_ENCODE: return (xdr_opaque(xdrs, sp, size)); case XDR_FREE: mem_free(sp, nodesize); *cpp = NULL; return (TRUE); } /* NOTREACHED */ return (FALSE); } /* * Wrapper for xdr_string that can be called directly from * routines like clnt_call */ bool_t -xdr_wrapstring(xdrs, cpp) - XDR *xdrs; - char **cpp; +xdr_wrapstring(XDR *xdrs, char **cpp) { return xdr_string(xdrs, cpp, LASTUNSIGNED); } /* * NOTE: xdr_hyper(), xdr_u_hyper(), xdr_longlong_t(), and xdr_u_longlong_t() * are in the "non-portable" section because they require that a `long long' * be a 64-bit type. * * --thorpej@netbsd.org, November 30, 1999 */ /* * XDR 64-bit integers */ bool_t -xdr_int64_t(xdrs, llp) - XDR *xdrs; - int64_t *llp; +xdr_int64_t(XDR *xdrs, int64_t *llp) { u_long ul[2]; switch (xdrs->x_op) { case XDR_ENCODE: ul[0] = (u_long)((u_int64_t)*llp >> 32) & 0xffffffff; ul[1] = (u_long)((u_int64_t)*llp) & 0xffffffff; if (XDR_PUTLONG(xdrs, (long *)&ul[0]) == FALSE) return (FALSE); return (XDR_PUTLONG(xdrs, (long *)&ul[1])); case XDR_DECODE: if (XDR_GETLONG(xdrs, (long *)&ul[0]) == FALSE) return (FALSE); if (XDR_GETLONG(xdrs, (long *)&ul[1]) == FALSE) return (FALSE); *llp = (int64_t) (((u_int64_t)ul[0] << 32) | ((u_int64_t)ul[1])); return (TRUE); case XDR_FREE: return (TRUE); } /* NOTREACHED */ return (FALSE); } /* * XDR unsigned 64-bit integers */ bool_t -xdr_u_int64_t(xdrs, ullp) - XDR *xdrs; - u_int64_t *ullp; +xdr_u_int64_t(XDR *xdrs, u_int64_t *ullp) { u_long ul[2]; switch (xdrs->x_op) { case XDR_ENCODE: ul[0] = (u_long)(*ullp >> 32) & 0xffffffff; ul[1] = (u_long)(*ullp) & 0xffffffff; if (XDR_PUTLONG(xdrs, (long *)&ul[0]) == FALSE) return (FALSE); return (XDR_PUTLONG(xdrs, (long *)&ul[1])); case XDR_DECODE: if (XDR_GETLONG(xdrs, (long *)&ul[0]) == FALSE) return (FALSE); if (XDR_GETLONG(xdrs, (long *)&ul[1]) == FALSE) return (FALSE); *ullp = (u_int64_t) (((u_int64_t)ul[0] << 32) | ((u_int64_t)ul[1])); return (TRUE); case XDR_FREE: return (TRUE); } /* NOTREACHED */ return (FALSE); } /* * XDR unsigned 64-bit integers */ bool_t -xdr_uint64_t(xdrs, ullp) - XDR *xdrs; - uint64_t *ullp; +xdr_uint64_t(XDR *xdrs, uint64_t *ullp) { u_long ul[2]; switch (xdrs->x_op) { case XDR_ENCODE: ul[0] = (u_long)(*ullp >> 32) & 0xffffffff; ul[1] = (u_long)(*ullp) & 0xffffffff; if (XDR_PUTLONG(xdrs, (long *)&ul[0]) == FALSE) return (FALSE); return (XDR_PUTLONG(xdrs, (long *)&ul[1])); case XDR_DECODE: if (XDR_GETLONG(xdrs, (long *)&ul[0]) == FALSE) return (FALSE); if (XDR_GETLONG(xdrs, (long *)&ul[1]) == FALSE) return (FALSE); *ullp = (u_int64_t) (((u_int64_t)ul[0] << 32) | ((u_int64_t)ul[1])); return (TRUE); case XDR_FREE: return (TRUE); } /* NOTREACHED */ return (FALSE); } /* * XDR hypers */ bool_t -xdr_hyper(xdrs, llp) - XDR *xdrs; - longlong_t *llp; +xdr_hyper(XDR *xdrs, longlong_t *llp) { /* * Don't bother open-coding this; it's a fair amount of code. Just * call xdr_int64_t(). */ return (xdr_int64_t(xdrs, (int64_t *)llp)); } /* * XDR unsigned hypers */ bool_t -xdr_u_hyper(xdrs, ullp) - XDR *xdrs; - u_longlong_t *ullp; +xdr_u_hyper(XDR *xdrs, u_longlong_t *ullp) { /* * Don't bother open-coding this; it's a fair amount of code. Just * call xdr_u_int64_t(). */ return (xdr_u_int64_t(xdrs, (u_int64_t *)ullp)); } /* * XDR longlong_t's */ bool_t -xdr_longlong_t(xdrs, llp) - XDR *xdrs; - longlong_t *llp; +xdr_longlong_t(XDR *xdrs, longlong_t *llp) { /* * Don't bother open-coding this; it's a fair amount of code. Just * call xdr_int64_t(). */ return (xdr_int64_t(xdrs, (int64_t *)llp)); } /* * XDR u_longlong_t's */ bool_t -xdr_u_longlong_t(xdrs, ullp) - XDR *xdrs; - u_longlong_t *ullp; +xdr_u_longlong_t(XDR *xdrs, u_longlong_t *ullp) { /* * Don't bother open-coding this; it's a fair amount of code. Just * call xdr_u_int64_t(). */ return (xdr_u_int64_t(xdrs, (u_int64_t *)ullp)); } Index: projects/clang-trunk/lib/libc/xdr/xdr_array.c =================================================================== --- projects/clang-trunk/lib/libc/xdr/xdr_array.c (revision 283870) +++ projects/clang-trunk/lib/libc/xdr/xdr_array.c (revision 283871) @@ -1,163 +1,160 @@ /* $NetBSD: xdr_array.c,v 1.12 2000/01/22 22:19:18 mycroft Exp $ */ /*- * Copyright (c) 2010, Oracle America, Inc. * * Redistribution and use in source and binary forms, with or without * modification, are permitted provided that the following conditions are * met: * * * Redistributions of source code must retain the above copyright * notice, this list of conditions and the following disclaimer. * * Redistributions in binary form must reproduce the above * copyright notice, this list of conditions and the following * disclaimer in the documentation and/or other materials * provided with the distribution. * * Neither the name of the "Oracle America, Inc." nor the names of its * contributors may be used to endorse or promote products derived * from this software without specific prior written permission. * * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS * FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE * COPYRIGHT HOLDER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, * INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE * GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, * WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING * NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE * OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. */ #if defined(LIBC_SCCS) && !defined(lint) static char *sccsid2 = "@(#)xdr_array.c 1.10 87/08/11 Copyr 1984 Sun Micro"; static char *sccsid = "@(#)xdr_array.c 2.1 88/07/29 4.0 RPCSRC"; #endif #include __FBSDID("$FreeBSD$"); /* * xdr_array.c, Generic XDR routines impelmentation. * * These are the "non-trivial" xdr primitives used to serialize and de-serialize * arrays. See xdr.h for more info on the interface to xdr. */ #include "namespace.h" #include #include #include #include #include #include #include #include "un-namespace.h" /* * XDR an array of arbitrary elements * *addrp is a pointer to the array, *sizep is the number of elements. * If addrp is NULL (*sizep * elsize) bytes are allocated. * elsize is the size (in bytes) of each element, and elproc is the * xdr procedure to call to handle each element of the array. */ bool_t -xdr_array(xdrs, addrp, sizep, maxsize, elsize, elproc) - XDR *xdrs; - caddr_t *addrp; /* array pointer */ - u_int *sizep; /* number of elements */ - u_int maxsize; /* max numberof elements */ - u_int elsize; /* size in bytes of each element */ - xdrproc_t elproc; /* xdr routine to handle each element */ +xdr_array(XDR *xdrs, caddr_t *addrp, u_int *sizep, u_int maxsize, u_int elsize, xdrproc_t elproc) +/* + * XDR *xdrs; + * caddr_t *addrp; // array pointer + * u_int *sizep; // number of elements + * u_int maxsize; // max numberof elements + * u_int elsize; // size in bytes of each element + * xdrproc_t elproc; // xdr routine to handle each element + */ { u_int i; caddr_t target = *addrp; u_int c; /* the actual element count */ bool_t stat = TRUE; u_int nodesize; /* like strings, arrays are really counted arrays */ if (!xdr_u_int(xdrs, sizep)) { return (FALSE); } c = *sizep; if ((c > maxsize || UINT_MAX/elsize < c) && (xdrs->x_op != XDR_FREE)) { return (FALSE); } nodesize = c * elsize; /* * if we are deserializing, we may need to allocate an array. * We also save time by checking for a null array if we are freeing. */ if (target == NULL) switch (xdrs->x_op) { case XDR_DECODE: if (c == 0) return (TRUE); *addrp = target = mem_alloc(nodesize); if (target == NULL) { warnx("xdr_array: out of memory"); return (FALSE); } memset(target, 0, nodesize); break; case XDR_FREE: return (TRUE); case XDR_ENCODE: break; } /* * now we xdr each element of array */ for (i = 0; (i < c) && stat; i++) { stat = (*elproc)(xdrs, target); target += elsize; } /* * the array may need freeing */ if (xdrs->x_op == XDR_FREE) { mem_free(*addrp, nodesize); *addrp = NULL; } return (stat); } /* * xdr_vector(): * * XDR a fixed length array. Unlike variable-length arrays, * the storage of fixed length arrays is static and unfreeable. * > basep: base of the array * > size: size of the array * > elemsize: size of each element * > xdr_elem: routine to XDR each element */ bool_t -xdr_vector(xdrs, basep, nelem, elemsize, xdr_elem) - XDR *xdrs; - char *basep; - u_int nelem; - u_int elemsize; - xdrproc_t xdr_elem; +xdr_vector(XDR *xdrs, char *basep, u_int nelem, u_int elemsize, xdrproc_t xdr_elem) { u_int i; char *elptr; elptr = basep; for (i = 0; i < nelem; i++) { if (!(*xdr_elem)(xdrs, elptr)) { return(FALSE); } elptr += elemsize; } return(TRUE); } Index: projects/clang-trunk/lib/libc/xdr/xdr_float.c =================================================================== --- projects/clang-trunk/lib/libc/xdr/xdr_float.c (revision 283870) +++ projects/clang-trunk/lib/libc/xdr/xdr_float.c (revision 283871) @@ -1,310 +1,308 @@ /* $NetBSD: xdr_float.c,v 1.23 2000/07/17 04:59:51 matt Exp $ */ /*- * Copyright (c) 2010, Oracle America, Inc. * * Redistribution and use in source and binary forms, with or without * modification, are permitted provided that the following conditions are * met: * * * Redistributions of source code must retain the above copyright * notice, this list of conditions and the following disclaimer. * * Redistributions in binary form must reproduce the above * copyright notice, this list of conditions and the following * disclaimer in the documentation and/or other materials * provided with the distribution. * * Neither the name of the "Oracle America, Inc." nor the names of its * contributors may be used to endorse or promote products derived * from this software without specific prior written permission. * * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS * FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE * COPYRIGHT HOLDER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, * INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE * GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, * WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING * NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE * OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. */ #if defined(LIBC_SCCS) && !defined(lint) static char *sccsid2 = "@(#)xdr_float.c 1.12 87/08/11 Copyr 1984 Sun Micro"; static char *sccsid = "@(#)xdr_float.c 2.1 88/07/29 4.0 RPCSRC"; #endif #include __FBSDID("$FreeBSD$"); /* * xdr_float.c, Generic XDR routines implementation. * * These are the "floating point" xdr routines used to (de)serialize * most common data items. See xdr.h for more info on the interface to * xdr. */ #include "namespace.h" #include #include #include #include #include #include "un-namespace.h" /* * NB: Not portable. * This routine works on machines with IEEE754 FP and Vaxen. */ #if defined(__m68k__) || defined(__sparc__) || defined(__i386__) || \ defined(__mips__) || defined(__ns32k__) || defined(__alpha__) || \ defined(__arm__) || defined(__ppc__) || \ defined(__arm26__) || defined(__sparc64__) || defined(__amd64__) || \ defined(__aarch64__) #include #define IEEEFP #endif #if defined(__vax__) /* What IEEE single precision floating point looks like on a Vax */ struct ieee_single { unsigned int mantissa: 23; unsigned int exp : 8; unsigned int sign : 1; }; /* Vax single precision floating point */ struct vax_single { unsigned int mantissa1 : 7; unsigned int exp : 8; unsigned int sign : 1; unsigned int mantissa2 : 16; }; #define VAX_SNG_BIAS 0x81 #define IEEE_SNG_BIAS 0x7f static struct sgl_limits { struct vax_single s; struct ieee_single ieee; } sgl_limits[2] = { {{ 0x7f, 0xff, 0x0, 0xffff }, /* Max Vax */ { 0x0, 0xff, 0x0 }}, /* Max IEEE */ {{ 0x0, 0x0, 0x0, 0x0 }, /* Min Vax */ { 0x0, 0x0, 0x0 }} /* Min IEEE */ }; #endif /* vax */ bool_t xdr_float(xdrs, fp) XDR *xdrs; float *fp; { #ifndef IEEEFP struct ieee_single is; struct vax_single vs, *vsp; struct sgl_limits *lim; int i; #endif switch (xdrs->x_op) { case XDR_ENCODE: #ifdef IEEEFP return (XDR_PUTINT32(xdrs, (int32_t *)fp)); #else vs = *((struct vax_single *)fp); for (i = 0, lim = sgl_limits; i < sizeof(sgl_limits)/sizeof(struct sgl_limits); i++, lim++) { if ((vs.mantissa2 == lim->s.mantissa2) && (vs.exp == lim->s.exp) && (vs.mantissa1 == lim->s.mantissa1)) { is = lim->ieee; goto shipit; } } is.exp = vs.exp - VAX_SNG_BIAS + IEEE_SNG_BIAS; is.mantissa = (vs.mantissa1 << 16) | vs.mantissa2; shipit: is.sign = vs.sign; return (XDR_PUTINT32(xdrs, (int32_t *)&is)); #endif case XDR_DECODE: #ifdef IEEEFP return (XDR_GETINT32(xdrs, (int32_t *)fp)); #else vsp = (struct vax_single *)fp; if (!XDR_GETINT32(xdrs, (int32_t *)&is)) return (FALSE); for (i = 0, lim = sgl_limits; i < sizeof(sgl_limits)/sizeof(struct sgl_limits); i++, lim++) { if ((is.exp == lim->ieee.exp) && (is.mantissa == lim->ieee.mantissa)) { *vsp = lim->s; goto doneit; } } vsp->exp = is.exp - IEEE_SNG_BIAS + VAX_SNG_BIAS; vsp->mantissa2 = is.mantissa; vsp->mantissa1 = (is.mantissa >> 16); doneit: vsp->sign = is.sign; return (TRUE); #endif case XDR_FREE: return (TRUE); } /* NOTREACHED */ return (FALSE); } #if defined(__vax__) /* What IEEE double precision floating point looks like on a Vax */ struct ieee_double { unsigned int mantissa1 : 20; unsigned int exp : 11; unsigned int sign : 1; unsigned int mantissa2 : 32; }; /* Vax double precision floating point */ struct vax_double { unsigned int mantissa1 : 7; unsigned int exp : 8; unsigned int sign : 1; unsigned int mantissa2 : 16; unsigned int mantissa3 : 16; unsigned int mantissa4 : 16; }; #define VAX_DBL_BIAS 0x81 #define IEEE_DBL_BIAS 0x3ff #define MASK(nbits) ((1 << nbits) - 1) static struct dbl_limits { struct vax_double d; struct ieee_double ieee; } dbl_limits[2] = { {{ 0x7f, 0xff, 0x0, 0xffff, 0xffff, 0xffff }, /* Max Vax */ { 0x0, 0x7ff, 0x0, 0x0 }}, /* Max IEEE */ {{ 0x0, 0x0, 0x0, 0x0, 0x0, 0x0}, /* Min Vax */ { 0x0, 0x0, 0x0, 0x0 }} /* Min IEEE */ }; #endif /* vax */ bool_t -xdr_double(xdrs, dp) - XDR *xdrs; - double *dp; +xdr_double(XDR *xdrs, double *dp) { #ifdef IEEEFP int32_t *i32p; bool_t rv; #else int32_t *lp; struct ieee_double id; struct vax_double vd; struct dbl_limits *lim; int i; #endif switch (xdrs->x_op) { case XDR_ENCODE: #ifdef IEEEFP i32p = (int32_t *)(void *)dp; #if BYTE_ORDER == BIG_ENDIAN rv = XDR_PUTINT32(xdrs, i32p); if (!rv) return (rv); rv = XDR_PUTINT32(xdrs, i32p+1); #else rv = XDR_PUTINT32(xdrs, i32p+1); if (!rv) return (rv); rv = XDR_PUTINT32(xdrs, i32p); #endif return (rv); #else vd = *((struct vax_double *)dp); for (i = 0, lim = dbl_limits; i < sizeof(dbl_limits)/sizeof(struct dbl_limits); i++, lim++) { if ((vd.mantissa4 == lim->d.mantissa4) && (vd.mantissa3 == lim->d.mantissa3) && (vd.mantissa2 == lim->d.mantissa2) && (vd.mantissa1 == lim->d.mantissa1) && (vd.exp == lim->d.exp)) { id = lim->ieee; goto shipit; } } id.exp = vd.exp - VAX_DBL_BIAS + IEEE_DBL_BIAS; id.mantissa1 = (vd.mantissa1 << 13) | (vd.mantissa2 >> 3); id.mantissa2 = ((vd.mantissa2 & MASK(3)) << 29) | (vd.mantissa3 << 13) | ((vd.mantissa4 >> 3) & MASK(13)); shipit: id.sign = vd.sign; lp = (int32_t *)&id; return (XDR_PUTINT32(xdrs, lp++) && XDR_PUTINT32(xdrs, lp)); #endif case XDR_DECODE: #ifdef IEEEFP i32p = (int32_t *)(void *)dp; #if BYTE_ORDER == BIG_ENDIAN rv = XDR_GETINT32(xdrs, i32p); if (!rv) return (rv); rv = XDR_GETINT32(xdrs, i32p+1); #else rv = XDR_GETINT32(xdrs, i32p+1); if (!rv) return (rv); rv = XDR_GETINT32(xdrs, i32p); #endif return (rv); #else lp = (int32_t *)&id; if (!XDR_GETINT32(xdrs, lp++) || !XDR_GETINT32(xdrs, lp)) return (FALSE); for (i = 0, lim = dbl_limits; i < sizeof(dbl_limits)/sizeof(struct dbl_limits); i++, lim++) { if ((id.mantissa2 == lim->ieee.mantissa2) && (id.mantissa1 == lim->ieee.mantissa1) && (id.exp == lim->ieee.exp)) { vd = lim->d; goto doneit; } } vd.exp = id.exp - IEEE_DBL_BIAS + VAX_DBL_BIAS; vd.mantissa1 = (id.mantissa1 >> 13); vd.mantissa2 = ((id.mantissa1 & MASK(13)) << 3) | (id.mantissa2 >> 29); vd.mantissa3 = (id.mantissa2 >> 13); vd.mantissa4 = (id.mantissa2 << 3); doneit: vd.sign = id.sign; *dp = *((double *)&vd); return (TRUE); #endif case XDR_FREE: return (TRUE); } /* NOTREACHED */ return (FALSE); } Index: projects/clang-trunk/lib/libc/xdr/xdr_mem.c =================================================================== --- projects/clang-trunk/lib/libc/xdr/xdr_mem.c (revision 283870) +++ projects/clang-trunk/lib/libc/xdr/xdr_mem.c (revision 283871) @@ -1,260 +1,234 @@ /* $NetBSD: xdr_mem.c,v 1.15 2000/01/22 22:19:18 mycroft Exp $ */ /*- * Copyright (c) 2010, Oracle America, Inc. * * Redistribution and use in source and binary forms, with or without * modification, are permitted provided that the following conditions are * met: * * * Redistributions of source code must retain the above copyright * notice, this list of conditions and the following disclaimer. * * Redistributions in binary form must reproduce the above * copyright notice, this list of conditions and the following * disclaimer in the documentation and/or other materials * provided with the distribution. * * Neither the name of the "Oracle America, Inc." nor the names of its * contributors may be used to endorse or promote products derived * from this software without specific prior written permission. * * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS * FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE * COPYRIGHT HOLDER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, * INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE * GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, * WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING * NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE * OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. */ #if defined(LIBC_SCCS) && !defined(lint) static char *sccsid2 = "@(#)xdr_mem.c 1.19 87/08/11 Copyr 1984 Sun Micro"; static char *sccsid = "@(#)xdr_mem.c 2.1 88/07/29 4.0 RPCSRC"; #endif #include __FBSDID("$FreeBSD$"); /* * xdr_mem.h, XDR implementation using memory buffers. * * If you have some data to be interpreted as external data representation * or to be converted to external data representation in a memory buffer, * then this is the package for you. * */ #include "namespace.h" #include #include #include #include #include #include "un-namespace.h" static void xdrmem_destroy(XDR *); static bool_t xdrmem_getlong_aligned(XDR *, long *); static bool_t xdrmem_putlong_aligned(XDR *, const long *); static bool_t xdrmem_getlong_unaligned(XDR *, long *); static bool_t xdrmem_putlong_unaligned(XDR *, const long *); static bool_t xdrmem_getbytes(XDR *, char *, u_int); static bool_t xdrmem_putbytes(XDR *, const char *, u_int); /* XXX: w/64-bit pointers, u_int not enough! */ static u_int xdrmem_getpos(XDR *); static bool_t xdrmem_setpos(XDR *, u_int); static int32_t *xdrmem_inline_aligned(XDR *, u_int); static int32_t *xdrmem_inline_unaligned(XDR *, u_int); static const struct xdr_ops xdrmem_ops_aligned = { xdrmem_getlong_aligned, xdrmem_putlong_aligned, xdrmem_getbytes, xdrmem_putbytes, xdrmem_getpos, xdrmem_setpos, xdrmem_inline_aligned, xdrmem_destroy }; static const struct xdr_ops xdrmem_ops_unaligned = { xdrmem_getlong_unaligned, xdrmem_putlong_unaligned, xdrmem_getbytes, xdrmem_putbytes, xdrmem_getpos, xdrmem_setpos, xdrmem_inline_unaligned, xdrmem_destroy }; /* * The procedure xdrmem_create initializes a stream descriptor for a * memory buffer. */ void -xdrmem_create(xdrs, addr, size, op) - XDR *xdrs; - char *addr; - u_int size; - enum xdr_op op; +xdrmem_create(XDR *xdrs, char *addr, u_int size, enum xdr_op op) { xdrs->x_op = op; xdrs->x_ops = ((unsigned long)addr & (sizeof(int32_t) - 1)) ? &xdrmem_ops_unaligned : &xdrmem_ops_aligned; xdrs->x_private = xdrs->x_base = addr; xdrs->x_handy = size; } /*ARGSUSED*/ static void -xdrmem_destroy(xdrs) - XDR *xdrs; +xdrmem_destroy(XDR *xdrs) { } static bool_t -xdrmem_getlong_aligned(xdrs, lp) - XDR *xdrs; - long *lp; +xdrmem_getlong_aligned(XDR *xdrs, long *lp) { if (xdrs->x_handy < sizeof(int32_t)) return (FALSE); xdrs->x_handy -= sizeof(int32_t); *lp = ntohl(*(u_int32_t *)xdrs->x_private); xdrs->x_private = (char *)xdrs->x_private + sizeof(int32_t); return (TRUE); } static bool_t -xdrmem_putlong_aligned(xdrs, lp) - XDR *xdrs; - const long *lp; +xdrmem_putlong_aligned(XDR *xdrs, const long *lp) { if (xdrs->x_handy < sizeof(int32_t)) return (FALSE); xdrs->x_handy -= sizeof(int32_t); *(u_int32_t *)xdrs->x_private = htonl((u_int32_t)*lp); xdrs->x_private = (char *)xdrs->x_private + sizeof(int32_t); return (TRUE); } static bool_t -xdrmem_getlong_unaligned(xdrs, lp) - XDR *xdrs; - long *lp; +xdrmem_getlong_unaligned(XDR *xdrs, long *lp) { u_int32_t l; if (xdrs->x_handy < sizeof(int32_t)) return (FALSE); xdrs->x_handy -= sizeof(int32_t); memmove(&l, xdrs->x_private, sizeof(int32_t)); *lp = ntohl(l); xdrs->x_private = (char *)xdrs->x_private + sizeof(int32_t); return (TRUE); } static bool_t -xdrmem_putlong_unaligned(xdrs, lp) - XDR *xdrs; - const long *lp; +xdrmem_putlong_unaligned(XDR *xdrs, const long *lp) { u_int32_t l; if (xdrs->x_handy < sizeof(int32_t)) return (FALSE); xdrs->x_handy -= sizeof(int32_t); l = htonl((u_int32_t)*lp); memmove(xdrs->x_private, &l, sizeof(int32_t)); xdrs->x_private = (char *)xdrs->x_private + sizeof(int32_t); return (TRUE); } static bool_t -xdrmem_getbytes(xdrs, addr, len) - XDR *xdrs; - char *addr; - u_int len; +xdrmem_getbytes(XDR *xdrs, char *addr, u_int len) { if (xdrs->x_handy < len) return (FALSE); xdrs->x_handy -= len; memmove(addr, xdrs->x_private, len); xdrs->x_private = (char *)xdrs->x_private + len; return (TRUE); } static bool_t -xdrmem_putbytes(xdrs, addr, len) - XDR *xdrs; - const char *addr; - u_int len; +xdrmem_putbytes(XDR *xdrs, const char *addr, u_int len) { if (xdrs->x_handy < len) return (FALSE); xdrs->x_handy -= len; memmove(xdrs->x_private, addr, len); xdrs->x_private = (char *)xdrs->x_private + len; return (TRUE); } static u_int -xdrmem_getpos(xdrs) - XDR *xdrs; +xdrmem_getpos(XDR *xdrs) { /* XXX w/64-bit pointers, u_int not enough! */ return (u_int)((u_long)xdrs->x_private - (u_long)xdrs->x_base); } static bool_t -xdrmem_setpos(xdrs, pos) - XDR *xdrs; - u_int pos; +xdrmem_setpos(XDR *xdrs, u_int pos) { char *newaddr = xdrs->x_base + pos; char *lastaddr = (char *)xdrs->x_private + xdrs->x_handy; if (newaddr > lastaddr) return (FALSE); xdrs->x_private = newaddr; xdrs->x_handy = (u_int)(lastaddr - newaddr); /* XXX sizeof(u_int) x_handy >= len) { xdrs->x_handy -= len; buf = (int32_t *)xdrs->x_private; xdrs->x_private = (char *)xdrs->x_private + len; } return (buf); } /* ARGSUSED */ static int32_t * -xdrmem_inline_unaligned(xdrs, len) - XDR *xdrs; - u_int len; +xdrmem_inline_unaligned(XDR *xdrs, u_int len) { return (0); } Index: projects/clang-trunk/lib/libc/xdr/xdr_rec.c =================================================================== --- projects/clang-trunk/lib/libc/xdr/xdr_rec.c (revision 283870) +++ projects/clang-trunk/lib/libc/xdr/xdr_rec.c (revision 283871) @@ -1,795 +1,761 @@ /* $NetBSD: xdr_rec.c,v 1.18 2000/07/06 03:10:35 christos Exp $ */ /*- * Copyright (c) 2010, Oracle America, Inc. * * Redistribution and use in source and binary forms, with or without * modification, are permitted provided that the following conditions are * met: * * * Redistributions of source code must retain the above copyright * notice, this list of conditions and the following disclaimer. * * Redistributions in binary form must reproduce the above * copyright notice, this list of conditions and the following * disclaimer in the documentation and/or other materials * provided with the distribution. * * Neither the name of the "Oracle America, Inc." nor the names of its * contributors may be used to endorse or promote products derived * from this software without specific prior written permission. * * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS * FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE * COPYRIGHT HOLDER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, * INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE * GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, * WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING * NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE * OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. */ #if defined(LIBC_SCCS) && !defined(lint) static char *sccsid2 = "@(#)xdr_rec.c 1.21 87/08/11 Copyr 1984 Sun Micro"; static char *sccsid = "@(#)xdr_rec.c 2.2 88/08/01 4.0 RPCSRC"; #endif #include __FBSDID("$FreeBSD$"); /* * xdr_rec.c, Implements TCP/IP based XDR streams with a "record marking" * layer above tcp (for rpc's use). * * These routines interface XDRSTREAMS to a tcp/ip connection. * There is a record marking layer between the xdr stream * and the tcp transport level. A record is composed on one or more * record fragments. A record fragment is a thirty-two bit header followed * by n bytes of data, where n is contained in the header. The header * is represented as a htonl(u_long). Thegh order bit encodes * whether or not the fragment is the last fragment of the record * (1 => fragment is last, 0 => more fragments to follow. * The other 31 bits encode the byte length of the fragment. */ #include "namespace.h" #include #include #include #include #include #include #include #include #include #include #include #include #include "un-namespace.h" #include "rpc_com.h" static bool_t xdrrec_getlong(XDR *, long *); static bool_t xdrrec_putlong(XDR *, const long *); static bool_t xdrrec_getbytes(XDR *, char *, u_int); static bool_t xdrrec_putbytes(XDR *, const char *, u_int); static u_int xdrrec_getpos(XDR *); static bool_t xdrrec_setpos(XDR *, u_int); static int32_t *xdrrec_inline(XDR *, u_int); static void xdrrec_destroy(XDR *); static const struct xdr_ops xdrrec_ops = { xdrrec_getlong, xdrrec_putlong, xdrrec_getbytes, xdrrec_putbytes, xdrrec_getpos, xdrrec_setpos, xdrrec_inline, xdrrec_destroy }; /* * A record is composed of one or more record fragments. * A record fragment is a four-byte header followed by zero to * 2**32-1 bytes. The header is treated as a long unsigned and is * encode/decoded to the network via htonl/ntohl. The low order 31 bits * are a byte count of the fragment. The highest order bit is a boolean: * 1 => this fragment is the last fragment of the record, * 0 => this fragment is followed by more fragment(s). * * The fragment/record machinery is not general; it is constructed to * meet the needs of xdr and rpc based on tcp. */ #define LAST_FRAG ((u_int32_t)(1U << 31)) typedef struct rec_strm { char *tcp_handle; /* * out-goung bits */ int (*writeit)(void *, void *, int); char *out_base; /* output buffer (points to frag header) */ char *out_finger; /* next output position */ char *out_boundry; /* data cannot up to this address */ u_int32_t *frag_header; /* beginning of curren fragment */ bool_t frag_sent; /* true if buffer sent in middle of record */ /* * in-coming bits */ int (*readit)(void *, void *, int); u_long in_size; /* fixed size of the input buffer */ char *in_base; char *in_finger; /* location of next byte to be had */ char *in_boundry; /* can read up to this location */ long fbtbc; /* fragment bytes to be consumed */ bool_t last_frag; u_int sendsize; u_int recvsize; bool_t nonblock; bool_t in_haveheader; u_int32_t in_header; char *in_hdrp; int in_hdrlen; int in_reclen; int in_received; int in_maxrec; } RECSTREAM; static u_int fix_buf_size(u_int); static bool_t flush_out(RECSTREAM *, bool_t); static bool_t fill_input_buf(RECSTREAM *); static bool_t get_input_bytes(RECSTREAM *, char *, int); static bool_t set_input_fragment(RECSTREAM *); static bool_t skip_input_bytes(RECSTREAM *, long); static bool_t realloc_stream(RECSTREAM *, int); /* * Create an xdr handle for xdrrec * xdrrec_create fills in xdrs. Sendsize and recvsize are * send and recv buffer sizes (0 => use default). * tcp_handle is an opaque handle that is passed as the first parameter to * the procedures readit and writeit. Readit and writeit are read and * write respectively. They are like the system * calls expect that they take an opaque handle rather than an fd. */ void -xdrrec_create(xdrs, sendsize, recvsize, tcp_handle, readit, writeit) - XDR *xdrs; - u_int sendsize; - u_int recvsize; - void *tcp_handle; - /* like read, but pass it a tcp_handle, not sock */ - int (*readit)(void *, void *, int); - /* like write, but pass it a tcp_handle, not sock */ - int (*writeit)(void *, void *, int); +xdrrec_create(XDR *xdrs, u_int sendsize, u_int recvsize, void *tcp_handle, + int (*readit)(void *, void *, int), int (*writeit)(void *, void *, int)) +/* + * XDR *xdrs; + * u_int sendsize; + * u_int recvsize; + * void *tcp_handle; + * // like read, but pass it a tcp_handle, not sock + * int (*readit)(void *, void *, int); + * // like write, but pass it a tcp_handle, not sock + * int (*writeit)(void *, void *, int); + */ { RECSTREAM *rstrm = mem_alloc(sizeof(RECSTREAM)); if (rstrm == NULL) { warnx("xdrrec_create: out of memory"); /* * This is bad. Should rework xdrrec_create to * return a handle, and in this case return NULL */ return; } rstrm->sendsize = sendsize = fix_buf_size(sendsize); rstrm->out_base = mem_alloc(rstrm->sendsize); if (rstrm->out_base == NULL) { warnx("xdrrec_create: out of memory"); mem_free(rstrm, sizeof(RECSTREAM)); return; } rstrm->recvsize = recvsize = fix_buf_size(recvsize); rstrm->in_base = mem_alloc(recvsize); if (rstrm->in_base == NULL) { warnx("xdrrec_create: out of memory"); mem_free(rstrm->out_base, sendsize); mem_free(rstrm, sizeof(RECSTREAM)); return; } /* * now the rest ... */ xdrs->x_ops = &xdrrec_ops; xdrs->x_private = rstrm; rstrm->tcp_handle = tcp_handle; rstrm->readit = readit; rstrm->writeit = writeit; rstrm->out_finger = rstrm->out_boundry = rstrm->out_base; rstrm->frag_header = (u_int32_t *)(void *)rstrm->out_base; rstrm->out_finger += sizeof(u_int32_t); rstrm->out_boundry += sendsize; rstrm->frag_sent = FALSE; rstrm->in_size = recvsize; rstrm->in_boundry = rstrm->in_base; rstrm->in_finger = (rstrm->in_boundry += recvsize); rstrm->fbtbc = 0; rstrm->last_frag = TRUE; rstrm->in_haveheader = FALSE; rstrm->in_hdrlen = 0; rstrm->in_hdrp = (char *)(void *)&rstrm->in_header; rstrm->nonblock = FALSE; rstrm->in_reclen = 0; rstrm->in_received = 0; } /* * The reoutines defined below are the xdr ops which will go into the * xdr handle filled in by xdrrec_create. */ static bool_t -xdrrec_getlong(xdrs, lp) - XDR *xdrs; - long *lp; +xdrrec_getlong(XDR *xdrs, long *lp) { RECSTREAM *rstrm = (RECSTREAM *)(xdrs->x_private); int32_t *buflp = (int32_t *)(void *)(rstrm->in_finger); int32_t mylong; /* first try the inline, fast case */ if ((rstrm->fbtbc >= sizeof(int32_t)) && (((long)rstrm->in_boundry - (long)buflp) >= sizeof(int32_t))) { *lp = (long)ntohl((u_int32_t)(*buflp)); rstrm->fbtbc -= sizeof(int32_t); rstrm->in_finger += sizeof(int32_t); } else { if (! xdrrec_getbytes(xdrs, (char *)(void *)&mylong, sizeof(int32_t))) return (FALSE); *lp = (long)ntohl((u_int32_t)mylong); } return (TRUE); } static bool_t -xdrrec_putlong(xdrs, lp) - XDR *xdrs; - const long *lp; +xdrrec_putlong(XDR *xdrs, const long *lp) { RECSTREAM *rstrm = (RECSTREAM *)(xdrs->x_private); int32_t *dest_lp = ((int32_t *)(void *)(rstrm->out_finger)); if ((rstrm->out_finger += sizeof(int32_t)) > rstrm->out_boundry) { /* * this case should almost never happen so the code is * inefficient */ rstrm->out_finger -= sizeof(int32_t); rstrm->frag_sent = TRUE; if (! flush_out(rstrm, FALSE)) return (FALSE); dest_lp = ((int32_t *)(void *)(rstrm->out_finger)); rstrm->out_finger += sizeof(int32_t); } *dest_lp = (int32_t)htonl((u_int32_t)(*lp)); return (TRUE); } static bool_t /* must manage buffers, fragments, and records */ -xdrrec_getbytes(xdrs, addr, len) - XDR *xdrs; - char *addr; - u_int len; +xdrrec_getbytes(XDR *xdrs, char *addr, u_int len) { RECSTREAM *rstrm = (RECSTREAM *)(xdrs->x_private); int current; while (len > 0) { current = (int)rstrm->fbtbc; if (current == 0) { if (rstrm->last_frag) return (FALSE); if (! set_input_fragment(rstrm)) return (FALSE); continue; } current = (len < current) ? len : current; if (! get_input_bytes(rstrm, addr, current)) return (FALSE); addr += current; rstrm->fbtbc -= current; len -= current; } return (TRUE); } static bool_t -xdrrec_putbytes(xdrs, addr, len) - XDR *xdrs; - const char *addr; - u_int len; +xdrrec_putbytes(XDR *xdrs, const char *addr, u_int len) { RECSTREAM *rstrm = (RECSTREAM *)(xdrs->x_private); size_t current; while (len > 0) { current = (size_t)((u_long)rstrm->out_boundry - (u_long)rstrm->out_finger); current = (len < current) ? len : current; memmove(rstrm->out_finger, addr, current); rstrm->out_finger += current; addr += current; len -= current; if (rstrm->out_finger == rstrm->out_boundry) { rstrm->frag_sent = TRUE; if (! flush_out(rstrm, FALSE)) return (FALSE); } } return (TRUE); } static u_int -xdrrec_getpos(xdrs) - XDR *xdrs; +xdrrec_getpos(XDR *xdrs) { RECSTREAM *rstrm = (RECSTREAM *)xdrs->x_private; off_t pos; pos = lseek((int)(u_long)rstrm->tcp_handle, (off_t)0, 1); if (pos == -1) pos = 0; switch (xdrs->x_op) { case XDR_ENCODE: pos += rstrm->out_finger - rstrm->out_base; break; case XDR_DECODE: pos -= rstrm->in_boundry - rstrm->in_finger; break; default: pos = (off_t) -1; break; } return ((u_int) pos); } static bool_t -xdrrec_setpos(xdrs, pos) - XDR *xdrs; - u_int pos; +xdrrec_setpos(XDR *xdrs, u_int pos) { RECSTREAM *rstrm = (RECSTREAM *)xdrs->x_private; u_int currpos = xdrrec_getpos(xdrs); int delta = currpos - pos; char *newpos; if ((int)currpos != -1) switch (xdrs->x_op) { case XDR_ENCODE: newpos = rstrm->out_finger - delta; if ((newpos > (char *)(void *)(rstrm->frag_header)) && (newpos < rstrm->out_boundry)) { rstrm->out_finger = newpos; return (TRUE); } break; case XDR_DECODE: newpos = rstrm->in_finger - delta; if ((delta < (int)(rstrm->fbtbc)) && (newpos <= rstrm->in_boundry) && (newpos >= rstrm->in_base)) { rstrm->in_finger = newpos; rstrm->fbtbc -= delta; return (TRUE); } break; case XDR_FREE: break; } return (FALSE); } static int32_t * -xdrrec_inline(xdrs, len) - XDR *xdrs; - u_int len; +xdrrec_inline(XDR *xdrs, u_int len) { RECSTREAM *rstrm = (RECSTREAM *)xdrs->x_private; int32_t *buf = NULL; switch (xdrs->x_op) { case XDR_ENCODE: if ((rstrm->out_finger + len) <= rstrm->out_boundry) { buf = (int32_t *)(void *)rstrm->out_finger; rstrm->out_finger += len; } break; case XDR_DECODE: if ((len <= rstrm->fbtbc) && ((rstrm->in_finger + len) <= rstrm->in_boundry)) { buf = (int32_t *)(void *)rstrm->in_finger; rstrm->fbtbc -= len; rstrm->in_finger += len; } break; case XDR_FREE: break; } return (buf); } static void -xdrrec_destroy(xdrs) - XDR *xdrs; +xdrrec_destroy(XDR *xdrs) { RECSTREAM *rstrm = (RECSTREAM *)xdrs->x_private; mem_free(rstrm->out_base, rstrm->sendsize); mem_free(rstrm->in_base, rstrm->recvsize); mem_free(rstrm, sizeof(RECSTREAM)); } /* * Exported routines to manage xdr records */ /* * Before reading (deserializing from the stream, one should always call * this procedure to guarantee proper record alignment. */ bool_t -xdrrec_skiprecord(xdrs) - XDR *xdrs; +xdrrec_skiprecord(XDR *xdrs) { RECSTREAM *rstrm = (RECSTREAM *)(xdrs->x_private); enum xprt_stat xstat; if (rstrm->nonblock) { if (__xdrrec_getrec(xdrs, &xstat, FALSE)) { rstrm->fbtbc = 0; return TRUE; } if (rstrm->in_finger == rstrm->in_boundry && xstat == XPRT_MOREREQS) { rstrm->fbtbc = 0; return TRUE; } return FALSE; } while (rstrm->fbtbc > 0 || (! rstrm->last_frag)) { if (! skip_input_bytes(rstrm, rstrm->fbtbc)) return (FALSE); rstrm->fbtbc = 0; if ((! rstrm->last_frag) && (! set_input_fragment(rstrm))) return (FALSE); } rstrm->last_frag = FALSE; return (TRUE); } /* * Look ahead function. * Returns TRUE iff there is no more input in the buffer * after consuming the rest of the current record. */ bool_t -xdrrec_eof(xdrs) - XDR *xdrs; +xdrrec_eof(XDR *xdrs) { RECSTREAM *rstrm = (RECSTREAM *)(xdrs->x_private); while (rstrm->fbtbc > 0 || (! rstrm->last_frag)) { if (! skip_input_bytes(rstrm, rstrm->fbtbc)) return (TRUE); rstrm->fbtbc = 0; if ((! rstrm->last_frag) && (! set_input_fragment(rstrm))) return (TRUE); } if (rstrm->in_finger == rstrm->in_boundry) return (TRUE); return (FALSE); } /* * The client must tell the package when an end-of-record has occurred. * The second paraemters tells whether the record should be flushed to the * (output) tcp stream. (This let's the package support batched or * pipelined procedure calls.) TRUE => immmediate flush to tcp connection. */ bool_t -xdrrec_endofrecord(xdrs, sendnow) - XDR *xdrs; - bool_t sendnow; +xdrrec_endofrecord(XDR *xdrs, bool_t sendnow) { RECSTREAM *rstrm = (RECSTREAM *)(xdrs->x_private); u_long len; /* fragment length */ if (sendnow || rstrm->frag_sent || ((u_long)rstrm->out_finger + sizeof(u_int32_t) >= (u_long)rstrm->out_boundry)) { rstrm->frag_sent = FALSE; return (flush_out(rstrm, TRUE)); } len = (u_long)(rstrm->out_finger) - (u_long)(rstrm->frag_header) - sizeof(u_int32_t); *(rstrm->frag_header) = htonl((u_int32_t)len | LAST_FRAG); rstrm->frag_header = (u_int32_t *)(void *)rstrm->out_finger; rstrm->out_finger += sizeof(u_int32_t); return (TRUE); } /* * Fill the stream buffer with a record for a non-blocking connection. * Return true if a record is available in the buffer, false if not. */ bool_t -__xdrrec_getrec(xdrs, statp, expectdata) - XDR *xdrs; - enum xprt_stat *statp; - bool_t expectdata; +__xdrrec_getrec(XDR *xdrs, enum xprt_stat *statp, bool_t expectdata) { RECSTREAM *rstrm = (RECSTREAM *)(xdrs->x_private); ssize_t n; int fraglen; if (!rstrm->in_haveheader) { n = rstrm->readit(rstrm->tcp_handle, rstrm->in_hdrp, (int)sizeof (rstrm->in_header) - rstrm->in_hdrlen); if (n == 0) { *statp = expectdata ? XPRT_DIED : XPRT_IDLE; return FALSE; } if (n < 0) { *statp = XPRT_DIED; return FALSE; } rstrm->in_hdrp += n; rstrm->in_hdrlen += n; if (rstrm->in_hdrlen < sizeof (rstrm->in_header)) { *statp = XPRT_MOREREQS; return FALSE; } rstrm->in_header = ntohl(rstrm->in_header); fraglen = (int)(rstrm->in_header & ~LAST_FRAG); if (fraglen == 0 || fraglen > rstrm->in_maxrec || (rstrm->in_reclen + fraglen) > rstrm->in_maxrec) { *statp = XPRT_DIED; return FALSE; } rstrm->in_reclen += fraglen; if (rstrm->in_reclen > rstrm->recvsize) realloc_stream(rstrm, rstrm->in_reclen); if (rstrm->in_header & LAST_FRAG) { rstrm->in_header &= ~LAST_FRAG; rstrm->last_frag = TRUE; } /* * We can only reasonably expect to read once from a * non-blocking stream. Reading the fragment header * may have drained the stream. */ expectdata = FALSE; } n = rstrm->readit(rstrm->tcp_handle, rstrm->in_base + rstrm->in_received, (rstrm->in_reclen - rstrm->in_received)); if (n < 0) { *statp = XPRT_DIED; return FALSE; } if (n == 0) { *statp = expectdata ? XPRT_DIED : XPRT_IDLE; return FALSE; } rstrm->in_received += n; if (rstrm->in_received == rstrm->in_reclen) { rstrm->in_haveheader = FALSE; rstrm->in_hdrp = (char *)(void *)&rstrm->in_header; rstrm->in_hdrlen = 0; if (rstrm->last_frag) { rstrm->fbtbc = rstrm->in_reclen; rstrm->in_boundry = rstrm->in_base + rstrm->in_reclen; rstrm->in_finger = rstrm->in_base; rstrm->in_reclen = rstrm->in_received = 0; *statp = XPRT_MOREREQS; return TRUE; } } *statp = XPRT_MOREREQS; return FALSE; } bool_t -__xdrrec_setnonblock(xdrs, maxrec) - XDR *xdrs; - int maxrec; +__xdrrec_setnonblock(XDR *xdrs, int maxrec) { RECSTREAM *rstrm = (RECSTREAM *)(xdrs->x_private); rstrm->nonblock = TRUE; if (maxrec == 0) maxrec = rstrm->recvsize; rstrm->in_maxrec = maxrec; return TRUE; } /* * Internal useful routines */ static bool_t -flush_out(rstrm, eor) - RECSTREAM *rstrm; - bool_t eor; +flush_out(RECSTREAM *rstrm, bool_t eor) { u_int32_t eormask = (eor == TRUE) ? LAST_FRAG : 0; u_int32_t len = (u_int32_t)((u_long)(rstrm->out_finger) - (u_long)(rstrm->frag_header) - sizeof(u_int32_t)); *(rstrm->frag_header) = htonl(len | eormask); len = (u_int32_t)((u_long)(rstrm->out_finger) - (u_long)(rstrm->out_base)); if ((*(rstrm->writeit))(rstrm->tcp_handle, rstrm->out_base, (int)len) != (int)len) return (FALSE); rstrm->frag_header = (u_int32_t *)(void *)rstrm->out_base; rstrm->out_finger = (char *)rstrm->out_base + sizeof(u_int32_t); return (TRUE); } static bool_t /* knows nothing about records! Only about input buffers */ -fill_input_buf(rstrm) - RECSTREAM *rstrm; +fill_input_buf(RECSTREAM *rstrm) { char *where; u_int32_t i; int len; if (rstrm->nonblock) return FALSE; where = rstrm->in_base; i = (u_int32_t)((u_long)rstrm->in_boundry % BYTES_PER_XDR_UNIT); where += i; len = (u_int32_t)(rstrm->in_size - i); if ((len = (*(rstrm->readit))(rstrm->tcp_handle, where, len)) == -1) return (FALSE); rstrm->in_finger = where; where += len; rstrm->in_boundry = where; return (TRUE); } static bool_t /* knows nothing about records! Only about input buffers */ -get_input_bytes(rstrm, addr, len) - RECSTREAM *rstrm; - char *addr; - int len; +get_input_bytes(RECSTREAM *rstrm, char *addr, int len) { size_t current; if (rstrm->nonblock) { if (len > (int)(rstrm->in_boundry - rstrm->in_finger)) return FALSE; memcpy(addr, rstrm->in_finger, (size_t)len); rstrm->in_finger += len; return TRUE; } while (len > 0) { current = (size_t)((long)rstrm->in_boundry - (long)rstrm->in_finger); if (current == 0) { if (! fill_input_buf(rstrm)) return (FALSE); continue; } current = (len < current) ? len : current; memmove(addr, rstrm->in_finger, current); rstrm->in_finger += current; addr += current; len -= current; } return (TRUE); } static bool_t /* next two bytes of the input stream are treated as a header */ -set_input_fragment(rstrm) - RECSTREAM *rstrm; +set_input_fragment(RECSTREAM *rstrm) { u_int32_t header; if (rstrm->nonblock) return FALSE; if (! get_input_bytes(rstrm, (char *)(void *)&header, sizeof(header))) return (FALSE); header = ntohl(header); rstrm->last_frag = ((header & LAST_FRAG) == 0) ? FALSE : TRUE; /* * Sanity check. Try not to accept wildly incorrect * record sizes. Unfortunately, the only record size * we can positively identify as being 'wildly incorrect' * is zero. Ridiculously large record sizes may look wrong, * but we don't have any way to be certain that they aren't * what the client actually intended to send us. */ if (header == 0) return(FALSE); rstrm->fbtbc = header & (~LAST_FRAG); return (TRUE); } static bool_t /* consumes input bytes; knows nothing about records! */ -skip_input_bytes(rstrm, cnt) - RECSTREAM *rstrm; - long cnt; +skip_input_bytes(RECSTREAM *rstrm, long cnt) { u_int32_t current; while (cnt > 0) { current = (size_t)((long)rstrm->in_boundry - (long)rstrm->in_finger); if (current == 0) { if (! fill_input_buf(rstrm)) return (FALSE); continue; } current = (u_int32_t)((cnt < current) ? cnt : current); rstrm->in_finger += current; cnt -= current; } return (TRUE); } static u_int -fix_buf_size(s) - u_int s; +fix_buf_size(u_int s) { if (s < 100) s = 4000; return (RNDUP(s)); } /* * Reallocate the input buffer for a non-block stream. */ static bool_t -realloc_stream(rstrm, size) - RECSTREAM *rstrm; - int size; +realloc_stream(RECSTREAM *rstrm, int size) { ptrdiff_t diff; char *buf; if (size > rstrm->recvsize) { buf = realloc(rstrm->in_base, (size_t)size); if (buf == NULL) return FALSE; diff = buf - rstrm->in_base; rstrm->in_finger += diff; rstrm->in_base = buf; rstrm->in_boundry = buf + size; rstrm->recvsize = size; rstrm->in_size = size; } return TRUE; } Index: projects/clang-trunk/lib/libc/xdr/xdr_reference.c =================================================================== --- projects/clang-trunk/lib/libc/xdr/xdr_reference.c (revision 283870) +++ projects/clang-trunk/lib/libc/xdr/xdr_reference.c (revision 283871) @@ -1,143 +1,141 @@ /* $NetBSD: xdr_reference.c,v 1.13 2000/01/22 22:19:18 mycroft Exp $ */ /*- * Copyright (c) 2010, Oracle America, Inc. * * Redistribution and use in source and binary forms, with or without * modification, are permitted provided that the following conditions are * met: * * * Redistributions of source code must retain the above copyright * notice, this list of conditions and the following disclaimer. * * Redistributions in binary form must reproduce the above * copyright notice, this list of conditions and the following * disclaimer in the documentation and/or other materials * provided with the distribution. * * Neither the name of the "Oracle America, Inc." nor the names of its * contributors may be used to endorse or promote products derived * from this software without specific prior written permission. * * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS * FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE * COPYRIGHT HOLDER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, * INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE * GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, * WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING * NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE * OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. */ #if defined(LIBC_SCCS) && !defined(lint) static char *sccsid2 = "@(#)xdr_reference.c 1.11 87/08/11 SMI"; static char *sccsid = "@(#)xdr_reference.c 2.1 88/07/29 4.0 RPCSRC"; #endif #include __FBSDID("$FreeBSD$"); /* * xdr_reference.c, Generic XDR routines impelmentation. * * These are the "non-trivial" xdr primitives used to serialize and de-serialize * "pointers". See xdr.h for more info on the interface to xdr. */ #include "namespace.h" #include #include #include #include #include #include #include "libc_private.h" /* * XDR an indirect pointer * xdr_reference is for recursively translating a structure that is * referenced by a pointer inside the structure that is currently being * translated. pp references a pointer to storage. If *pp is null * the necessary storage is allocated. * size is the sizeof the referneced structure. * proc is the routine to handle the referenced structure. */ bool_t -xdr_reference(xdrs, pp, size, proc) - XDR *xdrs; - caddr_t *pp; /* the pointer to work on */ - u_int size; /* size of the object pointed to */ - xdrproc_t proc; /* xdr routine to handle the object */ +xdr_reference(XDR *xdrs, caddr_t *pp, u_int size, xdrproc_t proc) +/* + * XDR *xdrs; + * caddr_t *pp; // the pointer to work on + * u_int size; // size of the object pointed to + * xdrproc_t proc; // xdr routine to handle the object + */ { caddr_t loc = *pp; bool_t stat; if (loc == NULL) switch (xdrs->x_op) { case XDR_FREE: return (TRUE); case XDR_DECODE: *pp = loc = (caddr_t) mem_alloc(size); if (loc == NULL) { warnx("xdr_reference: out of memory"); return (FALSE); } memset(loc, 0, size); break; case XDR_ENCODE: break; } stat = (*proc)(xdrs, loc); if (xdrs->x_op == XDR_FREE) { mem_free(loc, size); *pp = NULL; } return (stat); } /* * xdr_pointer(): * * XDR a pointer to a possibly recursive data structure. This * differs with xdr_reference in that it can serialize/deserialiaze * trees correctly. * * What's sent is actually a union: * * union object_pointer switch (boolean b) { * case TRUE: object_data data; * case FALSE: void nothing; * } * * > objpp: Pointer to the pointer to the object. * > obj_size: size of the object. * > xdr_obj: routine to XDR an object. * */ bool_t -xdr_pointer(xdrs,objpp,obj_size,xdr_obj) - XDR *xdrs; - char **objpp; - u_int obj_size; - xdrproc_t xdr_obj; +xdr_pointer(XDR *xdrs, char **objpp, u_int obj_size, xdrproc_t xdr_obj) { bool_t more_data; more_data = (*objpp != NULL); if (! xdr_bool(xdrs,&more_data)) { return (FALSE); } if (! more_data) { *objpp = NULL; return (TRUE); } return (xdr_reference(xdrs,objpp,obj_size,xdr_obj)); } Index: projects/clang-trunk/lib/libc/xdr/xdr_sizeof.c =================================================================== --- projects/clang-trunk/lib/libc/xdr/xdr_sizeof.c (revision 283870) +++ projects/clang-trunk/lib/libc/xdr/xdr_sizeof.c (revision 283871) @@ -1,168 +1,155 @@ /*- * Copyright (c) 2010, Oracle America, Inc. * * Redistribution and use in source and binary forms, with or without * modification, are permitted provided that the following conditions are * met: * * * Redistributions of source code must retain the above copyright * notice, this list of conditions and the following disclaimer. * * Redistributions in binary form must reproduce the above * copyright notice, this list of conditions and the following * disclaimer in the documentation and/or other materials * provided with the distribution. * * Neither the name of the "Oracle America, Inc." nor the names of its * contributors may be used to endorse or promote products derived * from this software without specific prior written permission. * * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS * FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE * COPYRIGHT HOLDER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, * INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE * GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, * WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING * NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE * OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. */ /* * xdr_sizeof.c * * General purpose routine to see how much space something will use * when serialized using XDR. */ #include __FBSDID("$FreeBSD$"); #include "namespace.h" #include #include #include #include #include "un-namespace.h" /* ARGSUSED */ static bool_t -x_putlong(xdrs, longp) - XDR *xdrs; - long *longp; +x_putlong(XDR *xdrs, const long *longp) { xdrs->x_handy += BYTES_PER_XDR_UNIT; return (TRUE); } /* ARGSUSED */ static bool_t -x_putbytes(xdrs, bp, len) - XDR *xdrs; - char *bp; - u_int len; +x_putbytes(XDR *xdrs, const char *bp, u_int len) { xdrs->x_handy += len; return (TRUE); } static u_int -x_getpostn(xdrs) - XDR *xdrs; +x_getpostn(XDR *xdrs) { return (xdrs->x_handy); } /* ARGSUSED */ static bool_t -x_setpostn(xdrs, pos) - XDR *xdrs; - u_int pos; +x_setpostn(XDR *xdrs, u_int pos) { /* This is not allowed */ return (FALSE); } static int32_t * -x_inline(xdrs, len) - XDR *xdrs; - u_int len; +x_inline(XDR *xdrs, u_int len) { if (len == 0) { return (NULL); } if (xdrs->x_op != XDR_ENCODE) { return (NULL); } if (len < (u_int)(uintptr_t)xdrs->x_base) { /* x_private was already allocated */ xdrs->x_handy += len; return ((int32_t *) xdrs->x_private); } else { /* Free the earlier space and allocate new area */ if (xdrs->x_private) free(xdrs->x_private); if ((xdrs->x_private = (caddr_t) malloc(len)) == NULL) { xdrs->x_base = 0; return (NULL); } xdrs->x_base = (caddr_t)(uintptr_t)len; xdrs->x_handy += len; return ((int32_t *) xdrs->x_private); } } static int -harmless() +harmless(void) { /* Always return FALSE/NULL, as the case may be */ return (0); } static void -x_destroy(xdrs) - XDR *xdrs; +x_destroy(XDR *xdrs) { xdrs->x_handy = 0; xdrs->x_base = 0; if (xdrs->x_private) { free(xdrs->x_private); xdrs->x_private = NULL; } return; } unsigned long -xdr_sizeof(func, data) - xdrproc_t func; - void *data; +xdr_sizeof(xdrproc_t func, void *data) { XDR x; struct xdr_ops ops; bool_t stat; /* to stop ANSI-C compiler from complaining */ typedef bool_t (* dummyfunc1)(XDR *, long *); typedef bool_t (* dummyfunc2)(XDR *, caddr_t, u_int); ops.x_putlong = x_putlong; ops.x_putbytes = x_putbytes; ops.x_inline = x_inline; ops.x_getpostn = x_getpostn; ops.x_setpostn = x_setpostn; ops.x_destroy = x_destroy; /* the other harmless ones */ ops.x_getlong = (dummyfunc1) harmless; ops.x_getbytes = (dummyfunc2) harmless; x.x_op = XDR_ENCODE; x.x_ops = &ops; x.x_handy = 0; x.x_private = (caddr_t) NULL; x.x_base = (caddr_t) 0; stat = func(&x, data); if (x.x_private) free(x.x_private); return (stat == TRUE ? (unsigned) x.x_handy: 0); } Index: projects/clang-trunk/lib/libc/xdr/xdr_stdio.c =================================================================== --- projects/clang-trunk/lib/libc/xdr/xdr_stdio.c (revision 283870) +++ projects/clang-trunk/lib/libc/xdr/xdr_stdio.c (revision 283871) @@ -1,196 +1,177 @@ /* $NetBSD: xdr_stdio.c,v 1.14 2000/01/22 22:19:19 mycroft Exp $ */ /*- * Copyright (c) 2010, Oracle America, Inc. * * Redistribution and use in source and binary forms, with or without * modification, are permitted provided that the following conditions are * met: * * * Redistributions of source code must retain the above copyright * notice, this list of conditions and the following disclaimer. * * Redistributions in binary form must reproduce the above * copyright notice, this list of conditions and the following * disclaimer in the documentation and/or other materials * provided with the distribution. * * Neither the name of the "Oracle America, Inc." nor the names of its * contributors may be used to endorse or promote products derived * from this software without specific prior written permission. * * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS * FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE * COPYRIGHT HOLDER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, * INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE * GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, * WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING * NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE * OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. */ #if defined(LIBC_SCCS) && !defined(lint) static char *sccsid2 = "@(#)xdr_stdio.c 1.16 87/08/11 Copyr 1984 Sun Micro"; static char *sccsid = "@(#)xdr_stdio.c 2.1 88/07/29 4.0 RPCSRC"; #endif #include __FBSDID("$FreeBSD$"); /* * xdr_stdio.c, XDR implementation on standard i/o file. * * This set of routines implements a XDR on a stdio stream. * XDR_ENCODE serializes onto the stream, XDR_DECODE de-serializes * from the stream. */ #include "namespace.h" #include #include #include #include #include "un-namespace.h" static void xdrstdio_destroy(XDR *); static bool_t xdrstdio_getlong(XDR *, long *); static bool_t xdrstdio_putlong(XDR *, const long *); static bool_t xdrstdio_getbytes(XDR *, char *, u_int); static bool_t xdrstdio_putbytes(XDR *, const char *, u_int); static u_int xdrstdio_getpos(XDR *); static bool_t xdrstdio_setpos(XDR *, u_int); static int32_t *xdrstdio_inline(XDR *, u_int); /* * Ops vector for stdio type XDR */ static const struct xdr_ops xdrstdio_ops = { xdrstdio_getlong, /* deseraialize a long int */ xdrstdio_putlong, /* seraialize a long int */ xdrstdio_getbytes, /* deserialize counted bytes */ xdrstdio_putbytes, /* serialize counted bytes */ xdrstdio_getpos, /* get offset in the stream */ xdrstdio_setpos, /* set offset in the stream */ xdrstdio_inline, /* prime stream for inline macros */ xdrstdio_destroy /* destroy stream */ }; /* * Initialize a stdio xdr stream. * Sets the xdr stream handle xdrs for use on the stream file. * Operation flag is set to op. */ void -xdrstdio_create(xdrs, file, op) - XDR *xdrs; - FILE *file; - enum xdr_op op; +xdrstdio_create(XDR *xdrs, FILE *file, enum xdr_op op) { xdrs->x_op = op; xdrs->x_ops = &xdrstdio_ops; xdrs->x_private = file; xdrs->x_handy = 0; xdrs->x_base = 0; } /* * Destroy a stdio xdr stream. * Cleans up the xdr stream handle xdrs previously set up by xdrstdio_create. */ static void -xdrstdio_destroy(xdrs) - XDR *xdrs; +xdrstdio_destroy(XDR *xdrs) { (void)fflush((FILE *)xdrs->x_private); /* XXX: should we close the file ?? */ } static bool_t -xdrstdio_getlong(xdrs, lp) - XDR *xdrs; - long *lp; +xdrstdio_getlong(XDR *xdrs, long *lp) { u_int32_t temp; if (fread(&temp, sizeof(int32_t), 1, (FILE *)xdrs->x_private) != 1) return (FALSE); *lp = (long)ntohl(temp); return (TRUE); } static bool_t -xdrstdio_putlong(xdrs, lp) - XDR *xdrs; - const long *lp; +xdrstdio_putlong(XDR *xdrs, const long *lp) { int32_t mycopy = htonl((u_int32_t)*lp); if (fwrite(&mycopy, sizeof(int32_t), 1, (FILE *)xdrs->x_private) != 1) return (FALSE); return (TRUE); } static bool_t -xdrstdio_getbytes(xdrs, addr, len) - XDR *xdrs; - char *addr; - u_int len; +xdrstdio_getbytes(XDR *xdrs, char *addr, u_int len) { if ((len != 0) && (fread(addr, (size_t)len, 1, (FILE *)xdrs->x_private) != 1)) return (FALSE); return (TRUE); } static bool_t -xdrstdio_putbytes(xdrs, addr, len) - XDR *xdrs; - const char *addr; - u_int len; +xdrstdio_putbytes(XDR *xdrs, const char *addr, u_int len) { if ((len != 0) && (fwrite(addr, (size_t)len, 1, (FILE *)xdrs->x_private) != 1)) return (FALSE); return (TRUE); } static u_int -xdrstdio_getpos(xdrs) - XDR *xdrs; +xdrstdio_getpos(XDR *xdrs) { return ((u_int) ftell((FILE *)xdrs->x_private)); } static bool_t -xdrstdio_setpos(xdrs, pos) - XDR *xdrs; - u_int pos; +xdrstdio_setpos(XDR *xdrs, u_int pos) { return ((fseek((FILE *)xdrs->x_private, (long)pos, 0) < 0) ? FALSE : TRUE); } /* ARGSUSED */ static int32_t * -xdrstdio_inline(xdrs, len) - XDR *xdrs; - u_int len; +xdrstdio_inline(XDR *xdrs, u_int len) { /* * Must do some work to implement this: must insure * enough data in the underlying stdio buffer, * that the buffer is aligned so that we can indirect through a * long *, and stuff this pointer in xdrs->x_buf. Doing * a fread or fwrite to a scratch buffer would defeat * most of the gains to be had here and require storage * management on this buffer, so we don't do this. */ return (NULL); } Index: projects/clang-trunk/lib/libc =================================================================== --- projects/clang-trunk/lib/libc (revision 283870) +++ projects/clang-trunk/lib/libc (revision 283871) Property changes on: projects/clang-trunk/lib/libc ___________________________________________________________________ Modified: svn:mergeinfo ## -0,0 +0,1 ## Merged /head/lib/libc:r283771-283870 Index: projects/clang-trunk/lib/libdpv/dpv.c =================================================================== --- projects/clang-trunk/lib/libdpv/dpv.c (revision 283870) +++ projects/clang-trunk/lib/libdpv/dpv.c (revision 283871) @@ -1,721 +1,721 @@ /*- * Copyright (c) 2013-2014 Devin Teske * All rights reserved. * * Redistribution and use in source and binary forms, with or without * modification, are permitted provided that the following conditions * are met: * 1. Redistributions of source code must retain the above copyright * notice, this list of conditions and the following disclaimer. * 2. Redistributions in binary form must reproduce the above copyright * notice, this list of conditions and the following disclaimer in the * documentation and/or other materials provided with the distribution. * * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE * ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF * SUCH DAMAGE. */ #include __FBSDID("$FreeBSD$"); #include #include #include #include #include #include #include #include #include #include #include #include #include #include "dialog_util.h" #include "dialogrc.h" #include "dprompt.h" #include "dpv.h" #include "dpv_private.h" #include "status.h" #include "util.h" /* Test Mechanics (Only used when dpv_config.options |= DPV_TEST_MODE) */ #define INCREMENT 1 /* Increment % per-pass test-mode */ #define XDIALOG_INCREMENT 15 /* different for slower Xdialog(1) */ static uint8_t increment = INCREMENT; /* Debugging */ uint8_t debug = FALSE; /* Data to process */ int dpv_interrupt = FALSE; int dpv_abort = FALSE; unsigned int dpv_nfiles = 0; /* Data processing */ long long dpv_overall_read = 0; static char pathbuf[PATH_MAX]; /* Extra display information */ uint8_t no_labels = FALSE; /* dpv_config.options & DPV_NO_LABELS */ uint8_t wide = FALSE; /* dpv_config.options & DPV_WIDE_MODE */ char *aprompt = NULL; /* dpv_config.aprompt */ char *msg_done = NULL; /* dpv_config.msg_done */ char *msg_fail = NULL; /* dpv_config.msg_fail */ char *msg_pending = NULL; /* dpv_config.msg_pending */ char *pprompt = NULL; /* dpv_config.pprompt */ /* Status-Line format for when using dialog(3) */ static const char *status_format_custom = NULL; static char status_format_default[DPV_STATUS_FORMAT_MAX]; /* * Takes a pointer to a dpv_config structure containing layout details and * pointer to initial element in a linked-list of dpv_file_node structures, * each presenting a file to process. Executes the `action' function passed-in * as a member to the `config' structure argument. */ int dpv(struct dpv_config *config, struct dpv_file_node *file_list) { char c; uint8_t keep_going; uint8_t nls = FALSE; /* See dialog_prompt_nlstate() */ uint8_t no_overrun = FALSE; uint8_t pprompt_nls = FALSE; /* See dialog_prompt_nlstate() */ uint8_t shrink_label_size = FALSE; mode_t mask; uint16_t options; char *cp; char *fc; char *last; char *name; char *output; const char *status_fmt; const char *path_fmt; enum dpv_display display_type; enum dpv_output output_type; enum dpv_status status; int (*action)(struct dpv_file_node *file, int out); int backslash; int dialog_last_update = 0; int dialog_old_nthfile = 0; int dialog_old_seconds = -1; int dialog_out = STDOUT_FILENO; int dialog_update_usec = 0; int dialog_updates_per_second; int files_left; int max_cols; int nthfile = 0; int output_out; int overall = 0; int pct; int res; int seconds; int status_last_update = 0; int status_old_nthfile = 0; int status_old_seconds = -1; int status_update_usec = 0; int status_updates_per_second; pid_t output_pid; pid_t pid; size_t len; struct dpv_file_node *curfile; struct dpv_file_node *first_file; struct dpv_file_node *list_head; struct timeval now; struct timeval start; char init_prompt[PROMPT_MAX + 1] = ""; /* Initialize globals to default values */ aprompt = NULL; pprompt = NULL; options = 0; action = NULL; backtitle = NULL; debug = FALSE; dialog_test = FALSE; dialog_updates_per_second = DIALOG_UPDATES_PER_SEC; display_limit = DISPLAY_LIMIT_DEFAULT; display_type = DPV_DISPLAY_LIBDIALOG; label_size = LABEL_SIZE_DEFAULT; msg_done = NULL; msg_fail = NULL; msg_pending = NULL; no_labels = FALSE; output = NULL; output_type = DPV_OUTPUT_NONE; pbar_size = PBAR_SIZE_DEFAULT; status_format_custom = NULL; status_updates_per_second = STATUS_UPDATES_PER_SEC; title = NULL; wide = FALSE; /* Process config options (overriding defaults) */ if (config != NULL) { if (config->aprompt != NULL) { if (aprompt == NULL) { aprompt = malloc(DPV_APROMPT_MAX); if (aprompt == NULL) return (-1); } snprintf(aprompt, DPV_APROMPT_MAX, "%s", config->aprompt); } if (config->pprompt != NULL) { if (pprompt == NULL) { pprompt = malloc(DPV_PPROMPT_MAX + 2); /* +2 is for implicit "\n" appended later */ if (pprompt == NULL) return (-1); } snprintf(pprompt, DPV_APROMPT_MAX, "%s", config->pprompt); } options = config->options; action = config->action; backtitle = config->backtitle; debug = config->debug; dialog_test = ((options & DPV_TEST_MODE) != 0); dialog_updates_per_second = config->dialog_updates_per_second; display_limit = config->display_limit; display_type = config->display_type; label_size = config->label_size; msg_done = (char *)config->msg_done; msg_fail = (char *)config->msg_fail; msg_pending = (char *)config->msg_pending; no_labels = ((options & DPV_NO_LABELS) != 0); no_overrun = ((options & DPV_NO_OVERRUN) != 0); output = config->output; output_type = config->output_type; pbar_size = config->pbar_size; status_updates_per_second = config->status_updates_per_second; title = config->title; wide = ((options & DPV_WIDE_MODE) != 0); /* Enforce some minimums (pedantic) */ if (display_limit < -1) display_limit = -1; if (label_size < -1) label_size = -1; if (pbar_size < -1) pbar_size = -1; /* For the mini-pbar, -1 means hide, zero is invalid unless * only one file is given */ if (pbar_size == 0) { if (file_list == NULL || file_list->next == NULL) pbar_size = -1; else pbar_size = PBAR_SIZE_DEFAULT; } /* For the label, -1 means auto-size, zero is invalid unless * specifically requested through the use of options flag */ if (label_size == 0 && no_labels == FALSE) label_size = LABEL_SIZE_DEFAULT; /* Status update should not be zero */ if (status_updates_per_second == 0) status_updates_per_second = STATUS_UPDATES_PER_SEC; } /* config != NULL */ /* Process the type of display we've been requested to produce */ switch (display_type) { case DPV_DISPLAY_STDOUT: debug = TRUE; use_color = FALSE; use_dialog = FALSE; use_libdialog = FALSE; use_xdialog = FALSE; break; case DPV_DISPLAY_DIALOG: use_color = TRUE; use_dialog = TRUE; use_libdialog = FALSE; use_xdialog = FALSE; break; case DPV_DISPLAY_XDIALOG: snprintf(dialog, PATH_MAX, XDIALOG); use_color = FALSE; use_dialog = FALSE; use_libdialog = FALSE; use_xdialog = TRUE; break; default: use_color = TRUE; use_dialog = FALSE; use_libdialog = TRUE; use_xdialog = FALSE; break; } /* display_type */ /* Enforce additional minimums that require knowing our display type */ if (dialog_updates_per_second == 0) dialog_updates_per_second = use_xdialog ? XDIALOG_UPDATES_PER_SEC : DIALOG_UPDATES_PER_SEC; /* Allow forceful override of use_color */ if (config != NULL && (config->options & DPV_USE_COLOR) != 0) use_color = TRUE; /* Count the number of files in provided list of dpv_file_node's */ if (use_dialog && pprompt != NULL && *pprompt != '\0') pprompt_nls = dialog_prompt_nlstate(pprompt); max_cols = dialog_maxcols(); if (label_size == -1) shrink_label_size = TRUE; /* Process file arguments */ for (curfile = file_list; curfile != NULL; curfile = curfile->next) { dpv_nfiles++; /* dialog(3) only expands literal newlines */ if (use_libdialog) strexpandnl(curfile->name); /* Optionally calculate label size for file */ if (shrink_label_size) { nls = FALSE; name = curfile->name; if (curfile == file_list) nls = pprompt_nls; last = (char *)dialog_prompt_lastline(name, nls); if (use_dialog) { c = *last; *last = '\0'; nls = dialog_prompt_nlstate(name); *last = c; } len = dialog_prompt_longestline(last, nls); if ((int)len > (label_size - 3)) { if (label_size > 0) label_size += 3; label_size = len; /* Room for ellipsis (unless NULL) */ if (label_size > 0) label_size += 3; } if (max_cols > 0 && label_size > (max_cols - pbar_size - 9)) label_size = max_cols - pbar_size - 9; } if (debug) warnx("label=[%s] path=[%s] size=%lli", curfile->name, curfile->path, curfile->length); } /* file_list */ /* Optionally process the contents of DIALOGRC (~/.dialogrc) */ if (use_dialog) { res = parse_dialogrc(); if (debug && res == 0) { warnx("Successfully read `%s' config file", DIALOGRC); warnx("use_shadow = %i (Boolean)", use_shadow); warnx("use_colors = %i (Boolean)", use_colors); warnx("gauge_color=[%s] (FBH)", gauge_color); } } else if (use_libdialog) { init_dialog(stdin, stdout); use_shadow = dialog_state.use_shadow; use_colors = dialog_state.use_colors; gauge_color[0] = 48 + dlg_color_table[GAUGE_ATTR].fg; gauge_color[1] = 48 + dlg_color_table[GAUGE_ATTR].bg; gauge_color[2] = dlg_color_table[GAUGE_ATTR].hilite ? 'b' : 'B'; gauge_color[3] = '\0'; end_dialog(); if (debug) { warnx("Finished initializing dialog(3) library"); warnx("use_shadow = %i (Boolean)", use_shadow); warnx("use_colors = %i (Boolean)", use_colors); warnx("gauge_color=[%s] (FBH)", gauge_color); } } /* Enable mini progress bar automatically for stdin streams if unable * to calculate progress (missing `lines:' syntax). */ if (dpv_nfiles <= 1 && file_list != NULL && file_list->length < 0 && !dialog_test) pbar_size = PBAR_SIZE_DEFAULT; /* If $USE_COLOR is set and non-NULL enable color; otherwise disable */ if ((cp = getenv(ENV_USE_COLOR)) != 0) use_color = *cp != '\0' ? 1 : 0; /* Print error and return `-1' if not given at least one name */ if (dpv_nfiles == 0) { warnx("%s: no labels provided", __func__); return (-1); } else if (debug) warnx("%s: %u label%s provided", __func__, dpv_nfiles, dpv_nfiles == 1 ? "" : "s"); /* If only one file and pbar size is zero, default to `-1' */ if (dpv_nfiles <= 1 && pbar_size == 0) pbar_size = -1; /* Print some debugging information */ if (debug) { warnx("%s: %s(%i) max rows x cols = %i x %i", __func__, use_xdialog ? XDIALOG : DIALOG, use_libdialog ? 3 : 1, dialog_maxrows(), dialog_maxcols()); } /* Xdialog(1) updates a lot slower than dialog(1) */ if (dialog_test && use_xdialog) increment = XDIALOG_INCREMENT; /* Always add implicit newline to pprompt (when specified) */ if (pprompt != NULL && *pprompt != '\0') { len = strlen(pprompt); /* * NOTE: pprompt = malloc(PPROMPT_MAX + 2) * NOTE: (see getopt(2) section above for pprompt allocation) */ pprompt[len++] = '\\'; pprompt[len++] = 'n'; pprompt[len++] = '\0'; } /* Xdialog(1) requires newlines (a) escaped and (b) in triplicate */ if (use_xdialog && pprompt != NULL) { /* Replace `\n' with `\n\\n\n' in pprompt */ len = strlen(pprompt); len += strcount(pprompt, "\\n") * 2; if (len > DPV_PPROMPT_MAX) errx(EXIT_FAILURE, "%s: Oops, pprompt buffer overflow " "(%zu > %i)", __func__, len, DPV_PPROMPT_MAX); if (replaceall(pprompt, "\\n", "\n\\n\n") < 0) err(EXIT_FAILURE, "%s: replaceall()", __func__); } /* libdialog requires literal newlines */ else if (use_libdialog && pprompt != NULL) strexpandnl(pprompt); /* Xdialog(1) requires newlines (a) escaped and (b) in triplicate */ if (use_xdialog && aprompt != NULL) { /* Replace `\n' with `\n\\n\n' in aprompt */ len = strlen(aprompt); len += strcount(aprompt, "\\n") * 2; if (len > DPV_APROMPT_MAX) errx(EXIT_FAILURE, "%s: Oops, aprompt buffer overflow " " (%zu > %i)", __func__, len, DPV_APROMPT_MAX); if (replaceall(aprompt, "\\n", "\n\\n\n") < 0) err(EXIT_FAILURE, "%s: replaceall()", __func__); } /* libdialog requires literal newlines */ else if (use_libdialog && aprompt != NULL) strexpandnl(aprompt); /* * Warn user about an obscure dialog(1) bug (neither Xdialog(1) nor * libdialog are affected) in the `--gauge' widget. If the first non- * whitespace letter of "{new_prompt}" in "XXX\n{new_prompt}\nXXX\n" * is a number, the number can sometimes be mistaken for a percentage * to the overall progressbar. Other nasty side-effects such as the * entire prompt not displaying or displaying improperly are caused by * this bug too. * * NOTE: When we can use color, we have a work-around... prefix the * output with `\Zn' (used to terminate ANSI and reset to normal). */ if (use_dialog && !use_color) { backslash = 0; /* First, check pprompt (falls through if NULL) */ fc = pprompt; while (fc != NULL && *fc != '\0') { if (*fc == '\n') /* leading literal newline OK */ break; if (!isspace(*fc) && *fc != '\\' && backslash == 0) break; else if (backslash > 0 && *fc != 'n') break; else if (*fc == '\\') { backslash++; if (backslash > 2) break; /* we're safe */ } fc++; } /* First non-whitespace character that dialog(1) will see */ if (fc != NULL && *fc >= '0' && *fc <= '9') warnx("%s: WARNING! text argument to `-p' begins with " "a number (not recommended)", __func__); else if (fc > pprompt) warnx("%s: WARNING! text argument to `-p' begins with " "whitespace (not recommended)", __func__); /* * If no pprompt or pprompt is all whitespace, check the first * file name provided to make sure it is alright too. */ if ((pprompt == NULL || *fc == '\0') && file_list != NULL) { first_file = file_list; fc = first_file->name; while (fc != NULL && *fc != '\0' && isspace(*fc)) fc++; /* First non-whitespace char that dialog(1) will see */ if (fc != NULL && *fc >= '0' && *fc <= '9') warnx("%s: WARNING! File name `%s' begins " "with a number (use `-p text' for safety)", __func__, first_file->name); } } dprompt_init(file_list); /* Reads: label_size pbar_size pprompt aprompt dpv_nfiles */ /* Inits: dheight and dwidth */ if (!debug) { /* Internally create the initial `--gauge' prompt text */ dprompt_recreate(file_list, (struct dpv_file_node *)NULL, 0); /* Spawn [X]dialog(1) `--gauge', returning pipe descriptor */ if (use_libdialog) { status_printf(""); dprompt_libprint(pprompt, aprompt, 0); } else { dprompt_sprint(init_prompt, pprompt, aprompt); dialog_out = dialog_spawn_gauge(init_prompt, &pid); dprompt_dprint(dialog_out, pprompt, aprompt, 0); } } /* !debug */ /* Seed the random(3) generator */ if (dialog_test) srandom(0xf1eeface); /* Set default/custom status line format */ if (dpv_nfiles > 1) { snprintf(status_format_default, DPV_STATUS_FORMAT_MAX, "%s", DPV_STATUS_MANY); status_format_custom = config->status_many; } else { snprintf(status_format_default, DPV_STATUS_FORMAT_MAX, "%s", DPV_STATUS_SOLO); status_format_custom = config->status_solo; } /* Add test mode identifier to default status line if enabled */ if (dialog_test && (strlen(status_format_default) + 12) < DPV_STATUS_FORMAT_MAX) strcat(status_format_default, " [TEST MODE]"); /* Verify custom status format */ status_fmt = fmtcheck(status_format_custom, status_format_default); if (status_format_custom != NULL && status_fmt == status_format_default) { warnx("WARNING! Invalid status_format configuration `%s'", status_format_custom); warnx("Default status_format `%s'", status_format_default); } /* Record when we started (used to prevent updating too quickly) */ (void)gettimeofday(&start, (struct timezone *)NULL); /* Calculate number of microseconds in-between sub-second updates */ if (status_updates_per_second != 0) status_update_usec = 1000000 / status_updates_per_second; if (dialog_updates_per_second != 0) dialog_update_usec = 1000000 / dialog_updates_per_second; /* * Process the file list [serially] (one for each argument passed) */ files_left = dpv_nfiles; list_head = file_list; for (curfile = file_list; curfile != NULL; curfile = curfile->next) { keep_going = TRUE; output_out = -1; pct = 0; nthfile++; files_left--; if (dpv_interrupt) break; if (dialog_test) pct = 0 - increment; /* Attempt to spawn output program for this file */ if (!dialog_test && output != NULL) { mask = umask(0022); (void)umask(mask); switch (output_type) { case DPV_OUTPUT_SHELL: output_out = shell_spawn_pipecmd(output, curfile->name, &output_pid); break; case DPV_OUTPUT_FILE: path_fmt = fmtcheck(output, "%s"); if (path_fmt == output) len = snprintf(pathbuf, PATH_MAX, output, curfile->name); else len = snprintf(pathbuf, PATH_MAX, "%s", output); if (len >= PATH_MAX) { warnx("%s:%d:%s: pathbuf[%u] too small" "to hold output argument", __FILE__, __LINE__, __func__, PATH_MAX); return (-1); } if ((output_out = open(pathbuf, O_CREAT|O_WRONLY, DEFFILEMODE & ~mask)) < 0) { warn("%s", pathbuf); return (-1); } break; default: break; } } while (!dpv_interrupt && keep_going) { if (dialog_test) { usleep(50000); pct += increment; dpv_overall_read += (int)(random() / 512 / dpv_nfiles); /* 512 limits fake readout to Megabytes */ } else if (action != NULL) pct = action(curfile, output_out); if (no_overrun || dialog_test) keep_going = (pct < 100); else { status = curfile->status; keep_going = (status == DPV_STATUS_RUNNING); } /* Get current time and calculate seconds elapsed */ gettimeofday(&now, (struct timezone *)NULL); now.tv_sec = now.tv_sec - start.tv_sec; now.tv_usec = now.tv_usec - start.tv_usec; if (now.tv_usec < 0) now.tv_sec--, now.tv_usec += 1000000; seconds = now.tv_sec + (now.tv_usec / 1000000.0); /* Update dialog (be it dialog(3), dialog(1), etc.) */ if ((dialog_updates_per_second != 0 && ( seconds != dialog_old_seconds || now.tv_usec - dialog_last_update >= dialog_update_usec || nthfile != dialog_old_nthfile )) || pct == 100 ) { /* Calculate overall progress (rounding up) */ overall = (100 * nthfile - 100 + pct) / dpv_nfiles; if (((100 * nthfile - 100 + pct) * 10 / dpv_nfiles % 100) > 50) overall++; dprompt_recreate(list_head, curfile, pct); if (use_libdialog && !debug) { /* Update dialog(3) widget */ dprompt_libprint(pprompt, aprompt, overall); } else { /* stdout, dialog(1), or Xdialog(1) */ dprompt_dprint(dialog_out, pprompt, aprompt, overall); fsync(dialog_out); } dialog_old_seconds = seconds; dialog_old_nthfile = nthfile; dialog_last_update = now.tv_usec; } /* Update the status line */ if ((use_libdialog && !debug) && status_updates_per_second != 0 && ( keep_going != TRUE || seconds != status_old_seconds || now.tv_usec - status_last_update >= status_update_usec || nthfile != status_old_nthfile ) ) { status_printf(status_fmt, dpv_overall_read, (dpv_overall_read / (seconds == 0 ? 1 : seconds) * 1.0), 1, /* XXX until we add parallelism XXX */ files_left); status_old_seconds = seconds; status_old_nthfile = nthfile; status_last_update = now.tv_usec; } } if (!dialog_test && output_out >= 0) { close(output_out); waitpid(output_pid, (int *)NULL, 0); } if (dpv_abort) break; /* Advance head of list when we hit the max display lines */ if (display_limit > 0 && nthfile % display_limit == 0) list_head = curfile->next; } if (!debug) { if (use_libdialog) end_dialog(); else { close(dialog_out); waitpid(pid, (int *)NULL, 0); } if (!dpv_interrupt) printf("\n"); } else - warnx("%s: %lli lines read", __func__, dpv_overall_read); + warnx("%s: %lli overall read", __func__, dpv_overall_read); if (dpv_interrupt || dpv_abort) return (-1); else return (0); } /* * Free allocated items initialized by dpv() */ void dpv_free(void) { dialogrc_free(); dprompt_free(); dialog_maxsize_free(); if (aprompt != NULL) { free(aprompt); aprompt = NULL; } if (pprompt != NULL) { free(pprompt); pprompt = NULL; } status_free(); } Index: projects/clang-trunk/lib/libprocstat/libprocstat.c =================================================================== --- projects/clang-trunk/lib/libprocstat/libprocstat.c (revision 283870) +++ projects/clang-trunk/lib/libprocstat/libprocstat.c (revision 283871) @@ -1,2529 +1,2529 @@ /*- * Copyright (c) 2009 Stanislav Sedov * Copyright (c) 1988, 1993 * The Regents of the University of California. All rights reserved. * * Redistribution and use in source and binary forms, with or without * modification, are permitted provided that the following conditions * are met: * 1. Redistributions of source code must retain the above copyright * notice, this list of conditions and the following disclaimer. * 2. Redistributions in binary form must reproduce the above copyright * notice, this list of conditions and the following disclaimer in the * documentation and/or other materials provided with the distribution. * 3. All advertising materials mentioning features or use of this software * must display the following acknowledgement: * This product includes software developed by the University of * California, Berkeley and its contributors. * 4. Neither the name of the University nor the names of its contributors * may be used to endorse or promote products derived from this software * without specific prior written permission. * * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF * SUCH DAMAGE. */ #include __FBSDID("$FreeBSD$"); #include #include #include #include #define _WANT_UCRED #include #undef _WANT_UCRED #include #include #include #include #include #include #include #include #include #include #include #include #include #include #define _WANT_FILE #include #include #include #include #include #define _KERNEL #include #include #include #include #include #include #undef _KERNEL #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include "libprocstat_internal.h" #include "common_kvm.h" #include "core.h" int statfs(const char *, struct statfs *); /* XXX */ #define PROCSTAT_KVM 1 #define PROCSTAT_SYSCTL 2 #define PROCSTAT_CORE 3 static char **getargv(struct procstat *procstat, struct kinfo_proc *kp, size_t nchr, int env); static char *getmnton(kvm_t *kd, struct mount *m); static struct kinfo_vmentry * kinfo_getvmmap_core(struct procstat_core *core, int *cntp); static Elf_Auxinfo *procstat_getauxv_core(struct procstat_core *core, unsigned int *cntp); static Elf_Auxinfo *procstat_getauxv_sysctl(pid_t pid, unsigned int *cntp); static struct filestat_list *procstat_getfiles_kvm( struct procstat *procstat, struct kinfo_proc *kp, int mmapped); static struct filestat_list *procstat_getfiles_sysctl( struct procstat *procstat, struct kinfo_proc *kp, int mmapped); static int procstat_get_pipe_info_sysctl(struct filestat *fst, struct pipestat *pipe, char *errbuf); static int procstat_get_pipe_info_kvm(kvm_t *kd, struct filestat *fst, struct pipestat *pipe, char *errbuf); static int procstat_get_pts_info_sysctl(struct filestat *fst, struct ptsstat *pts, char *errbuf); static int procstat_get_pts_info_kvm(kvm_t *kd, struct filestat *fst, struct ptsstat *pts, char *errbuf); static int procstat_get_sem_info_sysctl(struct filestat *fst, struct semstat *sem, char *errbuf); static int procstat_get_sem_info_kvm(kvm_t *kd, struct filestat *fst, struct semstat *sem, char *errbuf); static int procstat_get_shm_info_sysctl(struct filestat *fst, struct shmstat *shm, char *errbuf); static int procstat_get_shm_info_kvm(kvm_t *kd, struct filestat *fst, struct shmstat *shm, char *errbuf); static int procstat_get_socket_info_sysctl(struct filestat *fst, struct sockstat *sock, char *errbuf); static int procstat_get_socket_info_kvm(kvm_t *kd, struct filestat *fst, struct sockstat *sock, char *errbuf); static int to_filestat_flags(int flags); static int procstat_get_vnode_info_kvm(kvm_t *kd, struct filestat *fst, struct vnstat *vn, char *errbuf); static int procstat_get_vnode_info_sysctl(struct filestat *fst, struct vnstat *vn, char *errbuf); static gid_t *procstat_getgroups_core(struct procstat_core *core, unsigned int *count); static gid_t * procstat_getgroups_kvm(kvm_t *kd, struct kinfo_proc *kp, unsigned int *count); static gid_t *procstat_getgroups_sysctl(pid_t pid, unsigned int *count); static struct kinfo_kstack *procstat_getkstack_sysctl(pid_t pid, int *cntp); static int procstat_getosrel_core(struct procstat_core *core, int *osrelp); static int procstat_getosrel_kvm(kvm_t *kd, struct kinfo_proc *kp, int *osrelp); static int procstat_getosrel_sysctl(pid_t pid, int *osrelp); static int procstat_getpathname_core(struct procstat_core *core, char *pathname, size_t maxlen); static int procstat_getpathname_sysctl(pid_t pid, char *pathname, size_t maxlen); static int procstat_getrlimit_core(struct procstat_core *core, int which, struct rlimit* rlimit); static int procstat_getrlimit_kvm(kvm_t *kd, struct kinfo_proc *kp, int which, struct rlimit* rlimit); static int procstat_getrlimit_sysctl(pid_t pid, int which, struct rlimit* rlimit); static int procstat_getumask_core(struct procstat_core *core, unsigned short *maskp); static int procstat_getumask_kvm(kvm_t *kd, struct kinfo_proc *kp, unsigned short *maskp); static int procstat_getumask_sysctl(pid_t pid, unsigned short *maskp); static int vntype2psfsttype(int type); void procstat_close(struct procstat *procstat) { assert(procstat); if (procstat->type == PROCSTAT_KVM) kvm_close(procstat->kd); else if (procstat->type == PROCSTAT_CORE) procstat_core_close(procstat->core); procstat_freeargv(procstat); procstat_freeenvv(procstat); free(procstat); } struct procstat * procstat_open_sysctl(void) { struct procstat *procstat; procstat = calloc(1, sizeof(*procstat)); if (procstat == NULL) { warn("malloc()"); return (NULL); } procstat->type = PROCSTAT_SYSCTL; return (procstat); } struct procstat * procstat_open_kvm(const char *nlistf, const char *memf) { struct procstat *procstat; kvm_t *kd; char buf[_POSIX2_LINE_MAX]; procstat = calloc(1, sizeof(*procstat)); if (procstat == NULL) { warn("malloc()"); return (NULL); } kd = kvm_openfiles(nlistf, memf, NULL, O_RDONLY, buf); if (kd == NULL) { warnx("kvm_openfiles(): %s", buf); free(procstat); return (NULL); } procstat->type = PROCSTAT_KVM; procstat->kd = kd; return (procstat); } struct procstat * procstat_open_core(const char *filename) { struct procstat *procstat; struct procstat_core *core; procstat = calloc(1, sizeof(*procstat)); if (procstat == NULL) { warn("malloc()"); return (NULL); } core = procstat_core_open(filename); if (core == NULL) { free(procstat); return (NULL); } procstat->type = PROCSTAT_CORE; procstat->core = core; return (procstat); } struct kinfo_proc * procstat_getprocs(struct procstat *procstat, int what, int arg, unsigned int *count) { struct kinfo_proc *p0, *p; size_t len, olen; int name[4]; int cnt; int error; assert(procstat); assert(count); p = NULL; if (procstat->type == PROCSTAT_KVM) { *count = 0; p0 = kvm_getprocs(procstat->kd, what, arg, &cnt); if (p0 == NULL || cnt <= 0) return (NULL); *count = cnt; len = *count * sizeof(*p); p = malloc(len); if (p == NULL) { warnx("malloc(%zu)", len); goto fail; } bcopy(p0, p, len); return (p); } else if (procstat->type == PROCSTAT_SYSCTL) { len = 0; name[0] = CTL_KERN; name[1] = KERN_PROC; name[2] = what; name[3] = arg; error = sysctl(name, 4, NULL, &len, NULL, 0); if (error < 0 && errno != EPERM) { warn("sysctl(kern.proc)"); goto fail; } if (len == 0) { warnx("no processes?"); goto fail; } do { len += len / 10; p = reallocf(p, len); if (p == NULL) { warnx("reallocf(%zu)", len); goto fail; } olen = len; error = sysctl(name, 4, p, &len, NULL, 0); } while (error < 0 && errno == ENOMEM && olen == len); if (error < 0 && errno != EPERM) { warn("sysctl(kern.proc)"); goto fail; } /* Perform simple consistency checks. */ if ((len % sizeof(*p)) != 0 || p->ki_structsize != sizeof(*p)) { warnx("kinfo_proc structure size mismatch (len = %zu)", len); goto fail; } *count = len / sizeof(*p); return (p); } else if (procstat->type == PROCSTAT_CORE) { p = procstat_core_get(procstat->core, PSC_TYPE_PROC, NULL, &len); if ((len % sizeof(*p)) != 0 || p->ki_structsize != sizeof(*p)) { warnx("kinfo_proc structure size mismatch"); goto fail; } *count = len / sizeof(*p); return (p); } else { warnx("unknown access method: %d", procstat->type); return (NULL); } fail: if (p) free(p); return (NULL); } void procstat_freeprocs(struct procstat *procstat __unused, struct kinfo_proc *p) { if (p != NULL) free(p); p = NULL; } struct filestat_list * procstat_getfiles(struct procstat *procstat, struct kinfo_proc *kp, int mmapped) { switch(procstat->type) { case PROCSTAT_KVM: return (procstat_getfiles_kvm(procstat, kp, mmapped)); case PROCSTAT_SYSCTL: case PROCSTAT_CORE: return (procstat_getfiles_sysctl(procstat, kp, mmapped)); default: warnx("unknown access method: %d", procstat->type); return (NULL); } } void procstat_freefiles(struct procstat *procstat, struct filestat_list *head) { struct filestat *fst, *tmp; STAILQ_FOREACH_SAFE(fst, head, next, tmp) { if (fst->fs_path != NULL) free(fst->fs_path); free(fst); } free(head); if (procstat->vmentries != NULL) { free(procstat->vmentries); procstat->vmentries = NULL; } if (procstat->files != NULL) { free(procstat->files); procstat->files = NULL; } } static struct filestat * filestat_new_entry(void *typedep, int type, int fd, int fflags, int uflags, int refcount, off_t offset, char *path, cap_rights_t *cap_rightsp) { struct filestat *entry; entry = calloc(1, sizeof(*entry)); if (entry == NULL) { warn("malloc()"); return (NULL); } entry->fs_typedep = typedep; entry->fs_fflags = fflags; entry->fs_uflags = uflags; entry->fs_fd = fd; entry->fs_type = type; entry->fs_ref_count = refcount; entry->fs_offset = offset; entry->fs_path = path; if (cap_rightsp != NULL) entry->fs_cap_rights = *cap_rightsp; else cap_rights_init(&entry->fs_cap_rights); return (entry); } static struct vnode * getctty(kvm_t *kd, struct kinfo_proc *kp) { struct pgrp pgrp; struct proc proc; struct session sess; int error; assert(kp); error = kvm_read_all(kd, (unsigned long)kp->ki_paddr, &proc, sizeof(proc)); if (error == 0) { warnx("can't read proc struct at %p for pid %d", kp->ki_paddr, kp->ki_pid); return (NULL); } if (proc.p_pgrp == NULL) return (NULL); error = kvm_read_all(kd, (unsigned long)proc.p_pgrp, &pgrp, sizeof(pgrp)); if (error == 0) { warnx("can't read pgrp struct at %p for pid %d", proc.p_pgrp, kp->ki_pid); return (NULL); } error = kvm_read_all(kd, (unsigned long)pgrp.pg_session, &sess, sizeof(sess)); if (error == 0) { warnx("can't read session struct at %p for pid %d", pgrp.pg_session, kp->ki_pid); return (NULL); } return (sess.s_ttyvp); } static struct filestat_list * procstat_getfiles_kvm(struct procstat *procstat, struct kinfo_proc *kp, int mmapped) { struct file file; struct filedesc filed; struct vm_map_entry vmentry; struct vm_object object; struct vmspace vmspace; vm_map_entry_t entryp; vm_map_t map; vm_object_t objp; struct vnode *vp; struct file **ofiles; struct filestat *entry; struct filestat_list *head; kvm_t *kd; void *data; int i, fflags; int prot, type; unsigned int nfiles; assert(procstat); kd = procstat->kd; if (kd == NULL) return (NULL); if (kp->ki_fd == NULL) return (NULL); if (!kvm_read_all(kd, (unsigned long)kp->ki_fd, &filed, sizeof(filed))) { warnx("can't read filedesc at %p", (void *)kp->ki_fd); return (NULL); } /* * Allocate list head. */ head = malloc(sizeof(*head)); if (head == NULL) return (NULL); STAILQ_INIT(head); /* root directory vnode, if one. */ if (filed.fd_rdir) { entry = filestat_new_entry(filed.fd_rdir, PS_FST_TYPE_VNODE, -1, PS_FST_FFLAG_READ, PS_FST_UFLAG_RDIR, 0, 0, NULL, NULL); if (entry != NULL) STAILQ_INSERT_TAIL(head, entry, next); } /* current working directory vnode. */ if (filed.fd_cdir) { entry = filestat_new_entry(filed.fd_cdir, PS_FST_TYPE_VNODE, -1, PS_FST_FFLAG_READ, PS_FST_UFLAG_CDIR, 0, 0, NULL, NULL); if (entry != NULL) STAILQ_INSERT_TAIL(head, entry, next); } /* jail root, if any. */ if (filed.fd_jdir) { entry = filestat_new_entry(filed.fd_jdir, PS_FST_TYPE_VNODE, -1, PS_FST_FFLAG_READ, PS_FST_UFLAG_JAIL, 0, 0, NULL, NULL); if (entry != NULL) STAILQ_INSERT_TAIL(head, entry, next); } /* ktrace vnode, if one */ if (kp->ki_tracep) { entry = filestat_new_entry(kp->ki_tracep, PS_FST_TYPE_VNODE, -1, PS_FST_FFLAG_READ | PS_FST_FFLAG_WRITE, PS_FST_UFLAG_TRACE, 0, 0, NULL, NULL); if (entry != NULL) STAILQ_INSERT_TAIL(head, entry, next); } /* text vnode, if one */ if (kp->ki_textvp) { entry = filestat_new_entry(kp->ki_textvp, PS_FST_TYPE_VNODE, -1, PS_FST_FFLAG_READ, PS_FST_UFLAG_TEXT, 0, 0, NULL, NULL); if (entry != NULL) STAILQ_INSERT_TAIL(head, entry, next); } /* Controlling terminal. */ if ((vp = getctty(kd, kp)) != NULL) { entry = filestat_new_entry(vp, PS_FST_TYPE_VNODE, -1, PS_FST_FFLAG_READ | PS_FST_FFLAG_WRITE, PS_FST_UFLAG_CTTY, 0, 0, NULL, NULL); if (entry != NULL) STAILQ_INSERT_TAIL(head, entry, next); } nfiles = filed.fd_lastfile + 1; ofiles = malloc(nfiles * sizeof(struct file *)); if (ofiles == NULL) { warn("malloc(%zu)", nfiles * sizeof(struct file *)); goto do_mmapped; } if (!kvm_read_all(kd, (unsigned long)filed.fd_ofiles, ofiles, nfiles * sizeof(struct file *))) { warnx("cannot read file structures at %p", (void *)filed.fd_ofiles); free(ofiles); goto do_mmapped; } for (i = 0; i <= filed.fd_lastfile; i++) { if (ofiles[i] == NULL) continue; if (!kvm_read_all(kd, (unsigned long)ofiles[i], &file, sizeof(struct file))) { warnx("can't read file %d at %p", i, (void *)ofiles[i]); continue; } switch (file.f_type) { case DTYPE_VNODE: type = PS_FST_TYPE_VNODE; data = file.f_vnode; break; case DTYPE_SOCKET: type = PS_FST_TYPE_SOCKET; data = file.f_data; break; case DTYPE_PIPE: type = PS_FST_TYPE_PIPE; data = file.f_data; break; case DTYPE_FIFO: type = PS_FST_TYPE_FIFO; data = file.f_vnode; break; #ifdef DTYPE_PTS case DTYPE_PTS: type = PS_FST_TYPE_PTS; data = file.f_data; break; #endif case DTYPE_SEM: type = PS_FST_TYPE_SEM; data = file.f_data; break; case DTYPE_SHM: type = PS_FST_TYPE_SHM; data = file.f_data; break; default: continue; } /* XXXRW: No capability rights support for kvm yet. */ entry = filestat_new_entry(data, type, i, to_filestat_flags(file.f_flag), 0, 0, 0, NULL, NULL); if (entry != NULL) STAILQ_INSERT_TAIL(head, entry, next); } free(ofiles); do_mmapped: /* * Process mmapped files if requested. */ if (mmapped) { if (!kvm_read_all(kd, (unsigned long)kp->ki_vmspace, &vmspace, sizeof(vmspace))) { warnx("can't read vmspace at %p", (void *)kp->ki_vmspace); goto exit; } map = &vmspace.vm_map; for (entryp = map->header.next; entryp != &kp->ki_vmspace->vm_map.header; entryp = vmentry.next) { if (!kvm_read_all(kd, (unsigned long)entryp, &vmentry, sizeof(vmentry))) { warnx("can't read vm_map_entry at %p", (void *)entryp); continue; } if (vmentry.eflags & MAP_ENTRY_IS_SUB_MAP) continue; if ((objp = vmentry.object.vm_object) == NULL) continue; for (; objp; objp = object.backing_object) { if (!kvm_read_all(kd, (unsigned long)objp, &object, sizeof(object))) { warnx("can't read vm_object at %p", (void *)objp); break; } } /* We want only vnode objects. */ if (object.type != OBJT_VNODE) continue; prot = vmentry.protection; fflags = 0; if (prot & VM_PROT_READ) fflags = PS_FST_FFLAG_READ; if ((vmentry.eflags & MAP_ENTRY_COW) == 0 && prot & VM_PROT_WRITE) fflags |= PS_FST_FFLAG_WRITE; /* * Create filestat entry. */ entry = filestat_new_entry(object.handle, PS_FST_TYPE_VNODE, -1, fflags, PS_FST_UFLAG_MMAP, 0, 0, NULL, NULL); if (entry != NULL) STAILQ_INSERT_TAIL(head, entry, next); } } exit: return (head); } /* * kinfo types to filestat translation. */ static int kinfo_type2fst(int kftype) { static struct { int kf_type; int fst_type; } kftypes2fst[] = { { KF_TYPE_CRYPTO, PS_FST_TYPE_CRYPTO }, { KF_TYPE_FIFO, PS_FST_TYPE_FIFO }, { KF_TYPE_KQUEUE, PS_FST_TYPE_KQUEUE }, { KF_TYPE_MQUEUE, PS_FST_TYPE_MQUEUE }, { KF_TYPE_NONE, PS_FST_TYPE_NONE }, { KF_TYPE_PIPE, PS_FST_TYPE_PIPE }, { KF_TYPE_PTS, PS_FST_TYPE_PTS }, { KF_TYPE_SEM, PS_FST_TYPE_SEM }, { KF_TYPE_SHM, PS_FST_TYPE_SHM }, { KF_TYPE_SOCKET, PS_FST_TYPE_SOCKET }, { KF_TYPE_VNODE, PS_FST_TYPE_VNODE }, { KF_TYPE_UNKNOWN, PS_FST_TYPE_UNKNOWN } }; #define NKFTYPES (sizeof(kftypes2fst) / sizeof(*kftypes2fst)) unsigned int i; for (i = 0; i < NKFTYPES; i++) if (kftypes2fst[i].kf_type == kftype) break; if (i == NKFTYPES) return (PS_FST_TYPE_UNKNOWN); return (kftypes2fst[i].fst_type); } /* * kinfo flags to filestat translation. */ static int kinfo_fflags2fst(int kfflags) { static struct { int kf_flag; int fst_flag; } kfflags2fst[] = { { KF_FLAG_APPEND, PS_FST_FFLAG_APPEND }, { KF_FLAG_ASYNC, PS_FST_FFLAG_ASYNC }, { KF_FLAG_CREAT, PS_FST_FFLAG_CREAT }, { KF_FLAG_DIRECT, PS_FST_FFLAG_DIRECT }, { KF_FLAG_EXCL, PS_FST_FFLAG_EXCL }, { KF_FLAG_EXEC, PS_FST_FFLAG_EXEC }, { KF_FLAG_EXLOCK, PS_FST_FFLAG_EXLOCK }, { KF_FLAG_FSYNC, PS_FST_FFLAG_SYNC }, { KF_FLAG_HASLOCK, PS_FST_FFLAG_HASLOCK }, { KF_FLAG_NOFOLLOW, PS_FST_FFLAG_NOFOLLOW }, { KF_FLAG_NONBLOCK, PS_FST_FFLAG_NONBLOCK }, { KF_FLAG_READ, PS_FST_FFLAG_READ }, { KF_FLAG_SHLOCK, PS_FST_FFLAG_SHLOCK }, { KF_FLAG_TRUNC, PS_FST_FFLAG_TRUNC }, { KF_FLAG_WRITE, PS_FST_FFLAG_WRITE } }; #define NKFFLAGS (sizeof(kfflags2fst) / sizeof(*kfflags2fst)) unsigned int i; int flags; flags = 0; for (i = 0; i < NKFFLAGS; i++) if ((kfflags & kfflags2fst[i].kf_flag) != 0) flags |= kfflags2fst[i].fst_flag; return (flags); } static int kinfo_uflags2fst(int fd) { switch (fd) { case KF_FD_TYPE_CTTY: return (PS_FST_UFLAG_CTTY); case KF_FD_TYPE_CWD: return (PS_FST_UFLAG_CDIR); case KF_FD_TYPE_JAIL: return (PS_FST_UFLAG_JAIL); case KF_FD_TYPE_TEXT: return (PS_FST_UFLAG_TEXT); case KF_FD_TYPE_TRACE: return (PS_FST_UFLAG_TRACE); case KF_FD_TYPE_ROOT: return (PS_FST_UFLAG_RDIR); } return (0); } static struct kinfo_file * kinfo_getfile_core(struct procstat_core *core, int *cntp) { int cnt; size_t len; char *buf, *bp, *eb; struct kinfo_file *kif, *kp, *kf; buf = procstat_core_get(core, PSC_TYPE_FILES, NULL, &len); if (buf == NULL) return (NULL); /* * XXXMG: The code below is just copy&past from libutil. * The code duplication can be avoided if libutil * is extended to provide something like: * struct kinfo_file *kinfo_getfile_from_buf(const char *buf, * size_t len, int *cntp); */ /* Pass 1: count items */ cnt = 0; bp = buf; eb = buf + len; while (bp < eb) { kf = (struct kinfo_file *)(uintptr_t)bp; bp += kf->kf_structsize; cnt++; } kif = calloc(cnt, sizeof(*kif)); if (kif == NULL) { free(buf); return (NULL); } bp = buf; eb = buf + len; kp = kif; /* Pass 2: unpack */ while (bp < eb) { kf = (struct kinfo_file *)(uintptr_t)bp; /* Copy/expand into pre-zeroed buffer */ memcpy(kp, kf, kf->kf_structsize); /* Advance to next packed record */ bp += kf->kf_structsize; /* Set field size to fixed length, advance */ kp->kf_structsize = sizeof(*kp); kp++; } free(buf); *cntp = cnt; return (kif); /* Caller must free() return value */ } static struct filestat_list * procstat_getfiles_sysctl(struct procstat *procstat, struct kinfo_proc *kp, int mmapped) { struct kinfo_file *kif, *files; struct kinfo_vmentry *kve, *vmentries; struct filestat_list *head; struct filestat *entry; char *path; off_t offset; int cnt, fd, fflags; int i, type, uflags; int refcount; cap_rights_t cap_rights; assert(kp); if (kp->ki_fd == NULL) return (NULL); switch(procstat->type) { case PROCSTAT_SYSCTL: files = kinfo_getfile(kp->ki_pid, &cnt); break; case PROCSTAT_CORE: files = kinfo_getfile_core(procstat->core, &cnt); break; default: assert(!"invalid type"); } if (files == NULL && errno != EPERM) { warn("kinfo_getfile()"); return (NULL); } procstat->files = files; /* * Allocate list head. */ head = malloc(sizeof(*head)); if (head == NULL) return (NULL); STAILQ_INIT(head); for (i = 0; i < cnt; i++) { kif = &files[i]; type = kinfo_type2fst(kif->kf_type); fd = kif->kf_fd >= 0 ? kif->kf_fd : -1; fflags = kinfo_fflags2fst(kif->kf_flags); uflags = kinfo_uflags2fst(kif->kf_fd); refcount = kif->kf_ref_count; offset = kif->kf_offset; if (*kif->kf_path != '\0') path = strdup(kif->kf_path); else path = NULL; cap_rights = kif->kf_cap_rights; /* * Create filestat entry. */ entry = filestat_new_entry(kif, type, fd, fflags, uflags, refcount, offset, path, &cap_rights); if (entry != NULL) STAILQ_INSERT_TAIL(head, entry, next); } if (mmapped != 0) { vmentries = procstat_getvmmap(procstat, kp, &cnt); procstat->vmentries = vmentries; if (vmentries == NULL || cnt == 0) goto fail; for (i = 0; i < cnt; i++) { kve = &vmentries[i]; if (kve->kve_type != KVME_TYPE_VNODE) continue; fflags = 0; if (kve->kve_protection & KVME_PROT_READ) fflags = PS_FST_FFLAG_READ; if ((kve->kve_flags & KVME_FLAG_COW) == 0 && kve->kve_protection & KVME_PROT_WRITE) fflags |= PS_FST_FFLAG_WRITE; offset = kve->kve_offset; refcount = kve->kve_ref_count; if (*kve->kve_path != '\0') path = strdup(kve->kve_path); else path = NULL; entry = filestat_new_entry(kve, PS_FST_TYPE_VNODE, -1, fflags, PS_FST_UFLAG_MMAP, refcount, offset, path, NULL); if (entry != NULL) STAILQ_INSERT_TAIL(head, entry, next); } } fail: return (head); } int procstat_get_pipe_info(struct procstat *procstat, struct filestat *fst, struct pipestat *ps, char *errbuf) { assert(ps); if (procstat->type == PROCSTAT_KVM) { return (procstat_get_pipe_info_kvm(procstat->kd, fst, ps, errbuf)); } else if (procstat->type == PROCSTAT_SYSCTL || procstat->type == PROCSTAT_CORE) { return (procstat_get_pipe_info_sysctl(fst, ps, errbuf)); } else { warnx("unknown access method: %d", procstat->type); if (errbuf != NULL) snprintf(errbuf, _POSIX2_LINE_MAX, "error"); return (1); } } static int procstat_get_pipe_info_kvm(kvm_t *kd, struct filestat *fst, struct pipestat *ps, char *errbuf) { struct pipe pi; void *pipep; assert(kd); assert(ps); assert(fst); bzero(ps, sizeof(*ps)); pipep = fst->fs_typedep; if (pipep == NULL) goto fail; if (!kvm_read_all(kd, (unsigned long)pipep, &pi, sizeof(struct pipe))) { warnx("can't read pipe at %p", (void *)pipep); goto fail; } ps->addr = (uintptr_t)pipep; ps->peer = (uintptr_t)pi.pipe_peer; ps->buffer_cnt = pi.pipe_buffer.cnt; return (0); fail: if (errbuf != NULL) snprintf(errbuf, _POSIX2_LINE_MAX, "error"); return (1); } static int procstat_get_pipe_info_sysctl(struct filestat *fst, struct pipestat *ps, char *errbuf __unused) { struct kinfo_file *kif; assert(ps); assert(fst); bzero(ps, sizeof(*ps)); kif = fst->fs_typedep; if (kif == NULL) return (1); ps->addr = kif->kf_un.kf_pipe.kf_pipe_addr; ps->peer = kif->kf_un.kf_pipe.kf_pipe_peer; ps->buffer_cnt = kif->kf_un.kf_pipe.kf_pipe_buffer_cnt; return (0); } int procstat_get_pts_info(struct procstat *procstat, struct filestat *fst, struct ptsstat *pts, char *errbuf) { assert(pts); if (procstat->type == PROCSTAT_KVM) { return (procstat_get_pts_info_kvm(procstat->kd, fst, pts, errbuf)); } else if (procstat->type == PROCSTAT_SYSCTL || procstat->type == PROCSTAT_CORE) { return (procstat_get_pts_info_sysctl(fst, pts, errbuf)); } else { warnx("unknown access method: %d", procstat->type); if (errbuf != NULL) snprintf(errbuf, _POSIX2_LINE_MAX, "error"); return (1); } } static int procstat_get_pts_info_kvm(kvm_t *kd, struct filestat *fst, struct ptsstat *pts, char *errbuf) { struct tty tty; void *ttyp; assert(kd); assert(pts); assert(fst); bzero(pts, sizeof(*pts)); ttyp = fst->fs_typedep; if (ttyp == NULL) goto fail; if (!kvm_read_all(kd, (unsigned long)ttyp, &tty, sizeof(struct tty))) { warnx("can't read tty at %p", (void *)ttyp); goto fail; } pts->dev = dev2udev(kd, tty.t_dev); (void)kdevtoname(kd, tty.t_dev, pts->devname); return (0); fail: if (errbuf != NULL) snprintf(errbuf, _POSIX2_LINE_MAX, "error"); return (1); } static int procstat_get_pts_info_sysctl(struct filestat *fst, struct ptsstat *pts, char *errbuf __unused) { struct kinfo_file *kif; assert(pts); assert(fst); bzero(pts, sizeof(*pts)); kif = fst->fs_typedep; if (kif == NULL) return (0); pts->dev = kif->kf_un.kf_pts.kf_pts_dev; strlcpy(pts->devname, kif->kf_path, sizeof(pts->devname)); return (0); } int procstat_get_sem_info(struct procstat *procstat, struct filestat *fst, struct semstat *sem, char *errbuf) { assert(sem); if (procstat->type == PROCSTAT_KVM) { return (procstat_get_sem_info_kvm(procstat->kd, fst, sem, errbuf)); } else if (procstat->type == PROCSTAT_SYSCTL || procstat->type == PROCSTAT_CORE) { return (procstat_get_sem_info_sysctl(fst, sem, errbuf)); } else { warnx("unknown access method: %d", procstat->type); if (errbuf != NULL) snprintf(errbuf, _POSIX2_LINE_MAX, "error"); return (1); } } static int procstat_get_sem_info_kvm(kvm_t *kd, struct filestat *fst, struct semstat *sem, char *errbuf) { struct ksem ksem; void *ksemp; char *path; int i; assert(kd); assert(sem); assert(fst); bzero(sem, sizeof(*sem)); ksemp = fst->fs_typedep; if (ksemp == NULL) goto fail; if (!kvm_read_all(kd, (unsigned long)ksemp, &ksem, sizeof(struct ksem))) { warnx("can't read ksem at %p", (void *)ksemp); goto fail; } sem->mode = S_IFREG | ksem.ks_mode; sem->value = ksem.ks_value; if (fst->fs_path == NULL && ksem.ks_path != NULL) { path = malloc(MAXPATHLEN); for (i = 0; i < MAXPATHLEN - 1; i++) { if (!kvm_read_all(kd, (unsigned long)ksem.ks_path + i, path + i, 1)) break; if (path[i] == '\0') break; } path[i] = '\0'; if (i == 0) free(path); else fst->fs_path = path; } return (0); fail: if (errbuf != NULL) snprintf(errbuf, _POSIX2_LINE_MAX, "error"); return (1); } static int procstat_get_sem_info_sysctl(struct filestat *fst, struct semstat *sem, char *errbuf __unused) { struct kinfo_file *kif; assert(sem); assert(fst); bzero(sem, sizeof(*sem)); kif = fst->fs_typedep; if (kif == NULL) return (0); sem->value = kif->kf_un.kf_sem.kf_sem_value; sem->mode = kif->kf_un.kf_sem.kf_sem_mode; return (0); } int procstat_get_shm_info(struct procstat *procstat, struct filestat *fst, struct shmstat *shm, char *errbuf) { assert(shm); if (procstat->type == PROCSTAT_KVM) { return (procstat_get_shm_info_kvm(procstat->kd, fst, shm, errbuf)); } else if (procstat->type == PROCSTAT_SYSCTL || procstat->type == PROCSTAT_CORE) { return (procstat_get_shm_info_sysctl(fst, shm, errbuf)); } else { warnx("unknown access method: %d", procstat->type); if (errbuf != NULL) snprintf(errbuf, _POSIX2_LINE_MAX, "error"); return (1); } } static int procstat_get_shm_info_kvm(kvm_t *kd, struct filestat *fst, struct shmstat *shm, char *errbuf) { struct shmfd shmfd; void *shmfdp; char *path; int i; assert(kd); assert(shm); assert(fst); bzero(shm, sizeof(*shm)); shmfdp = fst->fs_typedep; if (shmfdp == NULL) goto fail; if (!kvm_read_all(kd, (unsigned long)shmfdp, &shmfd, sizeof(struct shmfd))) { warnx("can't read shmfd at %p", (void *)shmfdp); goto fail; } shm->mode = S_IFREG | shmfd.shm_mode; shm->size = shmfd.shm_size; if (fst->fs_path == NULL && shmfd.shm_path != NULL) { path = malloc(MAXPATHLEN); for (i = 0; i < MAXPATHLEN - 1; i++) { if (!kvm_read_all(kd, (unsigned long)shmfd.shm_path + i, path + i, 1)) break; if (path[i] == '\0') break; } path[i] = '\0'; if (i == 0) free(path); else fst->fs_path = path; } return (0); fail: if (errbuf != NULL) snprintf(errbuf, _POSIX2_LINE_MAX, "error"); return (1); } static int procstat_get_shm_info_sysctl(struct filestat *fst, struct shmstat *shm, char *errbuf __unused) { struct kinfo_file *kif; assert(shm); assert(fst); bzero(shm, sizeof(*shm)); kif = fst->fs_typedep; if (kif == NULL) return (0); shm->size = kif->kf_un.kf_file.kf_file_size; shm->mode = kif->kf_un.kf_file.kf_file_mode; return (0); } int procstat_get_vnode_info(struct procstat *procstat, struct filestat *fst, struct vnstat *vn, char *errbuf) { assert(vn); if (procstat->type == PROCSTAT_KVM) { return (procstat_get_vnode_info_kvm(procstat->kd, fst, vn, errbuf)); } else if (procstat->type == PROCSTAT_SYSCTL || procstat->type == PROCSTAT_CORE) { return (procstat_get_vnode_info_sysctl(fst, vn, errbuf)); } else { warnx("unknown access method: %d", procstat->type); if (errbuf != NULL) snprintf(errbuf, _POSIX2_LINE_MAX, "error"); return (1); } } static int procstat_get_vnode_info_kvm(kvm_t *kd, struct filestat *fst, struct vnstat *vn, char *errbuf) { /* Filesystem specific handlers. */ #define FSTYPE(fst) {#fst, fst##_filestat} struct { const char *tag; int (*handler)(kvm_t *kd, struct vnode *vp, struct vnstat *vn); } fstypes[] = { FSTYPE(devfs), FSTYPE(isofs), FSTYPE(msdosfs), FSTYPE(nfs), FSTYPE(smbfs), FSTYPE(udf), FSTYPE(ufs), #ifdef LIBPROCSTAT_ZFS FSTYPE(zfs), #endif }; #define NTYPES (sizeof(fstypes) / sizeof(*fstypes)) struct vnode vnode; char tagstr[12]; void *vp; - int error, found; + int error; unsigned int i; assert(kd); assert(vn); assert(fst); vp = fst->fs_typedep; if (vp == NULL) goto fail; error = kvm_read_all(kd, (unsigned long)vp, &vnode, sizeof(vnode)); if (error == 0) { warnx("can't read vnode at %p", (void *)vp); goto fail; } bzero(vn, sizeof(*vn)); vn->vn_type = vntype2psfsttype(vnode.v_type); if (vnode.v_type == VNON || vnode.v_type == VBAD) return (0); error = kvm_read_all(kd, (unsigned long)vnode.v_tag, tagstr, sizeof(tagstr)); if (error == 0) { warnx("can't read v_tag at %p", (void *)vp); goto fail; } tagstr[sizeof(tagstr) - 1] = '\0'; /* * Find appropriate handler. */ - for (i = 0, found = 0; i < NTYPES; i++) + for (i = 0; i < NTYPES; i++) if (!strcmp(fstypes[i].tag, tagstr)) { if (fstypes[i].handler(kd, &vnode, vn) != 0) { goto fail; } break; } if (i == NTYPES) { if (errbuf != NULL) snprintf(errbuf, _POSIX2_LINE_MAX, "?(%s)", tagstr); return (1); } vn->vn_mntdir = getmnton(kd, vnode.v_mount); if ((vnode.v_type == VBLK || vnode.v_type == VCHR) && vnode.v_rdev != NULL){ vn->vn_dev = dev2udev(kd, vnode.v_rdev); (void)kdevtoname(kd, vnode.v_rdev, vn->vn_devname); } else { vn->vn_dev = -1; } return (0); fail: if (errbuf != NULL) snprintf(errbuf, _POSIX2_LINE_MAX, "error"); return (1); } /* * kinfo vnode type to filestat translation. */ static int kinfo_vtype2fst(int kfvtype) { static struct { int kf_vtype; int fst_vtype; } kfvtypes2fst[] = { { KF_VTYPE_VBAD, PS_FST_VTYPE_VBAD }, { KF_VTYPE_VBLK, PS_FST_VTYPE_VBLK }, { KF_VTYPE_VCHR, PS_FST_VTYPE_VCHR }, { KF_VTYPE_VDIR, PS_FST_VTYPE_VDIR }, { KF_VTYPE_VFIFO, PS_FST_VTYPE_VFIFO }, { KF_VTYPE_VLNK, PS_FST_VTYPE_VLNK }, { KF_VTYPE_VNON, PS_FST_VTYPE_VNON }, { KF_VTYPE_VREG, PS_FST_VTYPE_VREG }, { KF_VTYPE_VSOCK, PS_FST_VTYPE_VSOCK } }; #define NKFVTYPES (sizeof(kfvtypes2fst) / sizeof(*kfvtypes2fst)) unsigned int i; for (i = 0; i < NKFVTYPES; i++) if (kfvtypes2fst[i].kf_vtype == kfvtype) break; if (i == NKFVTYPES) return (PS_FST_VTYPE_UNKNOWN); return (kfvtypes2fst[i].fst_vtype); } static int procstat_get_vnode_info_sysctl(struct filestat *fst, struct vnstat *vn, char *errbuf) { struct statfs stbuf; struct kinfo_file *kif; struct kinfo_vmentry *kve; uint64_t fileid; uint64_t size; char *name, *path; uint32_t fsid; uint16_t mode; uint32_t rdev; int vntype; int status; assert(fst); assert(vn); bzero(vn, sizeof(*vn)); if (fst->fs_typedep == NULL) return (1); if (fst->fs_uflags & PS_FST_UFLAG_MMAP) { kve = fst->fs_typedep; fileid = kve->kve_vn_fileid; fsid = kve->kve_vn_fsid; mode = kve->kve_vn_mode; path = kve->kve_path; rdev = kve->kve_vn_rdev; size = kve->kve_vn_size; vntype = kinfo_vtype2fst(kve->kve_vn_type); status = kve->kve_status; } else { kif = fst->fs_typedep; fileid = kif->kf_un.kf_file.kf_file_fileid; fsid = kif->kf_un.kf_file.kf_file_fsid; mode = kif->kf_un.kf_file.kf_file_mode; path = kif->kf_path; rdev = kif->kf_un.kf_file.kf_file_rdev; size = kif->kf_un.kf_file.kf_file_size; vntype = kinfo_vtype2fst(kif->kf_vnode_type); status = kif->kf_status; } vn->vn_type = vntype; if (vntype == PS_FST_VTYPE_VNON || vntype == PS_FST_VTYPE_VBAD) return (0); if ((status & KF_ATTR_VALID) == 0) { if (errbuf != NULL) { snprintf(errbuf, _POSIX2_LINE_MAX, "? (no info available)"); } return (1); } if (path && *path) { statfs(path, &stbuf); vn->vn_mntdir = strdup(stbuf.f_mntonname); } else vn->vn_mntdir = strdup("-"); vn->vn_dev = rdev; if (vntype == PS_FST_VTYPE_VBLK) { name = devname(rdev, S_IFBLK); if (name != NULL) strlcpy(vn->vn_devname, name, sizeof(vn->vn_devname)); } else if (vntype == PS_FST_VTYPE_VCHR) { name = devname(vn->vn_dev, S_IFCHR); if (name != NULL) strlcpy(vn->vn_devname, name, sizeof(vn->vn_devname)); } vn->vn_fsid = fsid; vn->vn_fileid = fileid; vn->vn_size = size; vn->vn_mode = mode; return (0); } int procstat_get_socket_info(struct procstat *procstat, struct filestat *fst, struct sockstat *sock, char *errbuf) { assert(sock); if (procstat->type == PROCSTAT_KVM) { return (procstat_get_socket_info_kvm(procstat->kd, fst, sock, errbuf)); } else if (procstat->type == PROCSTAT_SYSCTL || procstat->type == PROCSTAT_CORE) { return (procstat_get_socket_info_sysctl(fst, sock, errbuf)); } else { warnx("unknown access method: %d", procstat->type); if (errbuf != NULL) snprintf(errbuf, _POSIX2_LINE_MAX, "error"); return (1); } } static int procstat_get_socket_info_kvm(kvm_t *kd, struct filestat *fst, struct sockstat *sock, char *errbuf) { struct domain dom; struct inpcb inpcb; struct protosw proto; struct socket s; struct unpcb unpcb; ssize_t len; void *so; assert(kd); assert(sock); assert(fst); bzero(sock, sizeof(*sock)); so = fst->fs_typedep; if (so == NULL) goto fail; sock->so_addr = (uintptr_t)so; /* fill in socket */ if (!kvm_read_all(kd, (unsigned long)so, &s, sizeof(struct socket))) { warnx("can't read sock at %p", (void *)so); goto fail; } /* fill in protosw entry */ if (!kvm_read_all(kd, (unsigned long)s.so_proto, &proto, sizeof(struct protosw))) { warnx("can't read protosw at %p", (void *)s.so_proto); goto fail; } /* fill in domain */ if (!kvm_read_all(kd, (unsigned long)proto.pr_domain, &dom, sizeof(struct domain))) { warnx("can't read domain at %p", (void *)proto.pr_domain); goto fail; } if ((len = kvm_read(kd, (unsigned long)dom.dom_name, sock->dname, sizeof(sock->dname) - 1)) < 0) { warnx("can't read domain name at %p", (void *)dom.dom_name); sock->dname[0] = '\0'; } else sock->dname[len] = '\0'; /* * Fill in known data. */ sock->type = s.so_type; sock->proto = proto.pr_protocol; sock->dom_family = dom.dom_family; sock->so_pcb = (uintptr_t)s.so_pcb; /* * Protocol specific data. */ switch(dom.dom_family) { case AF_INET: case AF_INET6: if (proto.pr_protocol == IPPROTO_TCP) { if (s.so_pcb) { if (kvm_read(kd, (u_long)s.so_pcb, (char *)&inpcb, sizeof(struct inpcb)) != sizeof(struct inpcb)) { warnx("can't read inpcb at %p", (void *)s.so_pcb); } else sock->inp_ppcb = (uintptr_t)inpcb.inp_ppcb; } } break; case AF_UNIX: if (s.so_pcb) { if (kvm_read(kd, (u_long)s.so_pcb, (char *)&unpcb, sizeof(struct unpcb)) != sizeof(struct unpcb)){ warnx("can't read unpcb at %p", (void *)s.so_pcb); } else if (unpcb.unp_conn) { sock->so_rcv_sb_state = s.so_rcv.sb_state; sock->so_snd_sb_state = s.so_snd.sb_state; sock->unp_conn = (uintptr_t)unpcb.unp_conn; } } break; default: break; } return (0); fail: if (errbuf != NULL) snprintf(errbuf, _POSIX2_LINE_MAX, "error"); return (1); } static int procstat_get_socket_info_sysctl(struct filestat *fst, struct sockstat *sock, char *errbuf __unused) { struct kinfo_file *kif; assert(sock); assert(fst); bzero(sock, sizeof(*sock)); kif = fst->fs_typedep; if (kif == NULL) return (0); /* * Fill in known data. */ sock->type = kif->kf_sock_type; sock->proto = kif->kf_sock_protocol; sock->dom_family = kif->kf_sock_domain; sock->so_pcb = kif->kf_un.kf_sock.kf_sock_pcb; strlcpy(sock->dname, kif->kf_path, sizeof(sock->dname)); bcopy(&kif->kf_sa_local, &sock->sa_local, kif->kf_sa_local.ss_len); bcopy(&kif->kf_sa_peer, &sock->sa_peer, kif->kf_sa_peer.ss_len); /* * Protocol specific data. */ switch(sock->dom_family) { case AF_INET: case AF_INET6: if (sock->proto == IPPROTO_TCP) sock->inp_ppcb = kif->kf_un.kf_sock.kf_sock_inpcb; break; case AF_UNIX: if (kif->kf_un.kf_sock.kf_sock_unpconn != 0) { sock->so_rcv_sb_state = kif->kf_un.kf_sock.kf_sock_rcv_sb_state; sock->so_snd_sb_state = kif->kf_un.kf_sock.kf_sock_snd_sb_state; sock->unp_conn = kif->kf_un.kf_sock.kf_sock_unpconn; } break; default: break; } return (0); } /* * Descriptor flags to filestat translation. */ static int to_filestat_flags(int flags) { static struct { int flag; int fst_flag; } fstflags[] = { { FREAD, PS_FST_FFLAG_READ }, { FWRITE, PS_FST_FFLAG_WRITE }, { O_APPEND, PS_FST_FFLAG_APPEND }, { O_ASYNC, PS_FST_FFLAG_ASYNC }, { O_CREAT, PS_FST_FFLAG_CREAT }, { O_DIRECT, PS_FST_FFLAG_DIRECT }, { O_EXCL, PS_FST_FFLAG_EXCL }, { O_EXEC, PS_FST_FFLAG_EXEC }, { O_EXLOCK, PS_FST_FFLAG_EXLOCK }, { O_NOFOLLOW, PS_FST_FFLAG_NOFOLLOW }, { O_NONBLOCK, PS_FST_FFLAG_NONBLOCK }, { O_SHLOCK, PS_FST_FFLAG_SHLOCK }, { O_SYNC, PS_FST_FFLAG_SYNC }, { O_TRUNC, PS_FST_FFLAG_TRUNC } }; #define NFSTFLAGS (sizeof(fstflags) / sizeof(*fstflags)) int fst_flags; unsigned int i; fst_flags = 0; for (i = 0; i < NFSTFLAGS; i++) if (flags & fstflags[i].flag) fst_flags |= fstflags[i].fst_flag; return (fst_flags); } /* * Vnode type to filestate translation. */ static int vntype2psfsttype(int type) { static struct { int vtype; int fst_vtype; } vt2fst[] = { { VBAD, PS_FST_VTYPE_VBAD }, { VBLK, PS_FST_VTYPE_VBLK }, { VCHR, PS_FST_VTYPE_VCHR }, { VDIR, PS_FST_VTYPE_VDIR }, { VFIFO, PS_FST_VTYPE_VFIFO }, { VLNK, PS_FST_VTYPE_VLNK }, { VNON, PS_FST_VTYPE_VNON }, { VREG, PS_FST_VTYPE_VREG }, { VSOCK, PS_FST_VTYPE_VSOCK } }; #define NVFTYPES (sizeof(vt2fst) / sizeof(*vt2fst)) unsigned int i, fst_type; fst_type = PS_FST_VTYPE_UNKNOWN; for (i = 0; i < NVFTYPES; i++) { if (type == vt2fst[i].vtype) { fst_type = vt2fst[i].fst_vtype; break; } } return (fst_type); } static char * getmnton(kvm_t *kd, struct mount *m) { struct mount mnt; static struct mtab { struct mtab *next; struct mount *m; char mntonname[MNAMELEN + 1]; } *mhead = NULL; struct mtab *mt; for (mt = mhead; mt != NULL; mt = mt->next) if (m == mt->m) return (mt->mntonname); if (!kvm_read_all(kd, (unsigned long)m, &mnt, sizeof(struct mount))) { warnx("can't read mount table at %p", (void *)m); return (NULL); } if ((mt = malloc(sizeof (struct mtab))) == NULL) err(1, NULL); mt->m = m; bcopy(&mnt.mnt_stat.f_mntonname[0], &mt->mntonname[0], MNAMELEN); mt->mntonname[MNAMELEN] = '\0'; mt->next = mhead; mhead = mt; return (mt->mntonname); } /* * Auxiliary structures and functions to get process environment or * command line arguments. */ struct argvec { char *buf; size_t bufsize; char **argv; size_t argc; }; static struct argvec * argvec_alloc(size_t bufsize) { struct argvec *av; av = malloc(sizeof(*av)); if (av == NULL) return (NULL); av->bufsize = bufsize; av->buf = malloc(av->bufsize); if (av->buf == NULL) { free(av); return (NULL); } av->argc = 32; av->argv = malloc(sizeof(char *) * av->argc); if (av->argv == NULL) { free(av->buf); free(av); return (NULL); } return av; } static void argvec_free(struct argvec * av) { free(av->argv); free(av->buf); free(av); } static char ** getargv(struct procstat *procstat, struct kinfo_proc *kp, size_t nchr, int env) { int error, name[4], argc, i; struct argvec *av, **avp; enum psc_type type; size_t len; char *p, **argv; assert(procstat); assert(kp); if (procstat->type == PROCSTAT_KVM) { warnx("can't use kvm access method"); return (NULL); } if (procstat->type != PROCSTAT_SYSCTL && procstat->type != PROCSTAT_CORE) { warnx("unknown access method: %d", procstat->type); return (NULL); } if (nchr == 0 || nchr > ARG_MAX) nchr = ARG_MAX; avp = (struct argvec **)(env ? &procstat->argv : &procstat->envv); av = *avp; if (av == NULL) { av = argvec_alloc(nchr); if (av == NULL) { warn("malloc(%zu)", nchr); return (NULL); } *avp = av; } else if (av->bufsize < nchr) { av->buf = reallocf(av->buf, nchr); if (av->buf == NULL) { warn("malloc(%zu)", nchr); return (NULL); } } if (procstat->type == PROCSTAT_SYSCTL) { name[0] = CTL_KERN; name[1] = KERN_PROC; name[2] = env ? KERN_PROC_ENV : KERN_PROC_ARGS; name[3] = kp->ki_pid; len = nchr; error = sysctl(name, 4, av->buf, &len, NULL, 0); if (error != 0 && errno != ESRCH && errno != EPERM) warn("sysctl(kern.proc.%s)", env ? "env" : "args"); if (error != 0 || len == 0) return (NULL); } else /* procstat->type == PROCSTAT_CORE */ { type = env ? PSC_TYPE_ENVV : PSC_TYPE_ARGV; len = nchr; if (procstat_core_get(procstat->core, type, av->buf, &len) == NULL) { return (NULL); } } argv = av->argv; argc = av->argc; i = 0; for (p = av->buf; p < av->buf + len; p += strlen(p) + 1) { argv[i++] = p; if (i < argc) continue; /* Grow argv. */ argc += argc; argv = realloc(argv, sizeof(char *) * argc); if (argv == NULL) { warn("malloc(%zu)", sizeof(char *) * argc); return (NULL); } av->argv = argv; av->argc = argc; } argv[i] = NULL; return (argv); } /* * Return process command line arguments. */ char ** procstat_getargv(struct procstat *procstat, struct kinfo_proc *p, size_t nchr) { return (getargv(procstat, p, nchr, 0)); } /* * Free the buffer allocated by procstat_getargv(). */ void procstat_freeargv(struct procstat *procstat) { if (procstat->argv != NULL) { argvec_free(procstat->argv); procstat->argv = NULL; } } /* * Return process environment. */ char ** procstat_getenvv(struct procstat *procstat, struct kinfo_proc *p, size_t nchr) { return (getargv(procstat, p, nchr, 1)); } /* * Free the buffer allocated by procstat_getenvv(). */ void procstat_freeenvv(struct procstat *procstat) { if (procstat->envv != NULL) { argvec_free(procstat->envv); procstat->envv = NULL; } } static struct kinfo_vmentry * kinfo_getvmmap_core(struct procstat_core *core, int *cntp) { int cnt; size_t len; char *buf, *bp, *eb; struct kinfo_vmentry *kiv, *kp, *kv; buf = procstat_core_get(core, PSC_TYPE_VMMAP, NULL, &len); if (buf == NULL) return (NULL); /* * XXXMG: The code below is just copy&past from libutil. * The code duplication can be avoided if libutil * is extended to provide something like: * struct kinfo_vmentry *kinfo_getvmmap_from_buf(const char *buf, * size_t len, int *cntp); */ /* Pass 1: count items */ cnt = 0; bp = buf; eb = buf + len; while (bp < eb) { kv = (struct kinfo_vmentry *)(uintptr_t)bp; bp += kv->kve_structsize; cnt++; } kiv = calloc(cnt, sizeof(*kiv)); if (kiv == NULL) { free(buf); return (NULL); } bp = buf; eb = buf + len; kp = kiv; /* Pass 2: unpack */ while (bp < eb) { kv = (struct kinfo_vmentry *)(uintptr_t)bp; /* Copy/expand into pre-zeroed buffer */ memcpy(kp, kv, kv->kve_structsize); /* Advance to next packed record */ bp += kv->kve_structsize; /* Set field size to fixed length, advance */ kp->kve_structsize = sizeof(*kp); kp++; } free(buf); *cntp = cnt; return (kiv); /* Caller must free() return value */ } struct kinfo_vmentry * procstat_getvmmap(struct procstat *procstat, struct kinfo_proc *kp, unsigned int *cntp) { switch(procstat->type) { case PROCSTAT_KVM: warnx("kvm method is not supported"); return (NULL); case PROCSTAT_SYSCTL: return (kinfo_getvmmap(kp->ki_pid, cntp)); case PROCSTAT_CORE: return (kinfo_getvmmap_core(procstat->core, cntp)); default: warnx("unknown access method: %d", procstat->type); return (NULL); } } void procstat_freevmmap(struct procstat *procstat __unused, struct kinfo_vmentry *vmmap) { free(vmmap); } static gid_t * procstat_getgroups_kvm(kvm_t *kd, struct kinfo_proc *kp, unsigned int *cntp) { struct proc proc; struct ucred ucred; gid_t *groups; size_t len; assert(kd != NULL); assert(kp != NULL); if (!kvm_read_all(kd, (unsigned long)kp->ki_paddr, &proc, sizeof(proc))) { warnx("can't read proc struct at %p for pid %d", kp->ki_paddr, kp->ki_pid); return (NULL); } if (proc.p_ucred == NOCRED) return (NULL); if (!kvm_read_all(kd, (unsigned long)proc.p_ucred, &ucred, sizeof(ucred))) { warnx("can't read ucred struct at %p for pid %d", proc.p_ucred, kp->ki_pid); return (NULL); } len = ucred.cr_ngroups * sizeof(gid_t); groups = malloc(len); if (groups == NULL) { warn("malloc(%zu)", len); return (NULL); } if (!kvm_read_all(kd, (unsigned long)ucred.cr_groups, groups, len)) { warnx("can't read groups at %p for pid %d", ucred.cr_groups, kp->ki_pid); free(groups); return (NULL); } *cntp = ucred.cr_ngroups; return (groups); } static gid_t * procstat_getgroups_sysctl(pid_t pid, unsigned int *cntp) { int mib[4]; size_t len; gid_t *groups; mib[0] = CTL_KERN; mib[1] = KERN_PROC; mib[2] = KERN_PROC_GROUPS; mib[3] = pid; len = (sysconf(_SC_NGROUPS_MAX) + 1) * sizeof(gid_t); groups = malloc(len); if (groups == NULL) { warn("malloc(%zu)", len); return (NULL); } if (sysctl(mib, 4, groups, &len, NULL, 0) == -1) { warn("sysctl: kern.proc.groups: %d", pid); free(groups); return (NULL); } *cntp = len / sizeof(gid_t); return (groups); } static gid_t * procstat_getgroups_core(struct procstat_core *core, unsigned int *cntp) { size_t len; gid_t *groups; groups = procstat_core_get(core, PSC_TYPE_GROUPS, NULL, &len); if (groups == NULL) return (NULL); *cntp = len / sizeof(gid_t); return (groups); } gid_t * procstat_getgroups(struct procstat *procstat, struct kinfo_proc *kp, unsigned int *cntp) { switch(procstat->type) { case PROCSTAT_KVM: return (procstat_getgroups_kvm(procstat->kd, kp, cntp)); case PROCSTAT_SYSCTL: return (procstat_getgroups_sysctl(kp->ki_pid, cntp)); case PROCSTAT_CORE: return (procstat_getgroups_core(procstat->core, cntp)); default: warnx("unknown access method: %d", procstat->type); return (NULL); } } void procstat_freegroups(struct procstat *procstat __unused, gid_t *groups) { free(groups); } static int procstat_getumask_kvm(kvm_t *kd, struct kinfo_proc *kp, unsigned short *maskp) { struct filedesc fd; assert(kd != NULL); assert(kp != NULL); if (kp->ki_fd == NULL) return (-1); if (!kvm_read_all(kd, (unsigned long)kp->ki_fd, &fd, sizeof(fd))) { warnx("can't read filedesc at %p for pid %d", kp->ki_fd, kp->ki_pid); return (-1); } *maskp = fd.fd_cmask; return (0); } static int procstat_getumask_sysctl(pid_t pid, unsigned short *maskp) { int error; int mib[4]; size_t len; mib[0] = CTL_KERN; mib[1] = KERN_PROC; mib[2] = KERN_PROC_UMASK; mib[3] = pid; len = sizeof(*maskp); error = sysctl(mib, 4, maskp, &len, NULL, 0); if (error != 0 && errno != ESRCH && errno != EPERM) warn("sysctl: kern.proc.umask: %d", pid); return (error); } static int procstat_getumask_core(struct procstat_core *core, unsigned short *maskp) { size_t len; unsigned short *buf; buf = procstat_core_get(core, PSC_TYPE_UMASK, NULL, &len); if (buf == NULL) return (-1); if (len < sizeof(*maskp)) { free(buf); return (-1); } *maskp = *buf; free(buf); return (0); } int procstat_getumask(struct procstat *procstat, struct kinfo_proc *kp, unsigned short *maskp) { switch(procstat->type) { case PROCSTAT_KVM: return (procstat_getumask_kvm(procstat->kd, kp, maskp)); case PROCSTAT_SYSCTL: return (procstat_getumask_sysctl(kp->ki_pid, maskp)); case PROCSTAT_CORE: return (procstat_getumask_core(procstat->core, maskp)); default: warnx("unknown access method: %d", procstat->type); return (-1); } } static int procstat_getrlimit_kvm(kvm_t *kd, struct kinfo_proc *kp, int which, struct rlimit* rlimit) { struct proc proc; unsigned long offset; assert(kd != NULL); assert(kp != NULL); assert(which >= 0 && which < RLIM_NLIMITS); if (!kvm_read_all(kd, (unsigned long)kp->ki_paddr, &proc, sizeof(proc))) { warnx("can't read proc struct at %p for pid %d", kp->ki_paddr, kp->ki_pid); return (-1); } if (proc.p_limit == NULL) return (-1); offset = (unsigned long)proc.p_limit + sizeof(struct rlimit) * which; if (!kvm_read_all(kd, offset, rlimit, sizeof(*rlimit))) { warnx("can't read rlimit struct at %p for pid %d", (void *)offset, kp->ki_pid); return (-1); } return (0); } static int procstat_getrlimit_sysctl(pid_t pid, int which, struct rlimit* rlimit) { int error, name[5]; size_t len; name[0] = CTL_KERN; name[1] = KERN_PROC; name[2] = KERN_PROC_RLIMIT; name[3] = pid; name[4] = which; len = sizeof(struct rlimit); error = sysctl(name, 5, rlimit, &len, NULL, 0); if (error < 0 && errno != ESRCH) { warn("sysctl: kern.proc.rlimit: %d", pid); return (-1); } if (error < 0 || len != sizeof(struct rlimit)) return (-1); return (0); } static int procstat_getrlimit_core(struct procstat_core *core, int which, struct rlimit* rlimit) { size_t len; struct rlimit* rlimits; if (which < 0 || which >= RLIM_NLIMITS) { errno = EINVAL; warn("getrlimit: which"); return (-1); } rlimits = procstat_core_get(core, PSC_TYPE_RLIMIT, NULL, &len); if (rlimits == NULL) return (-1); if (len < sizeof(struct rlimit) * RLIM_NLIMITS) { free(rlimits); return (-1); } *rlimit = rlimits[which]; return (0); } int procstat_getrlimit(struct procstat *procstat, struct kinfo_proc *kp, int which, struct rlimit* rlimit) { switch(procstat->type) { case PROCSTAT_KVM: return (procstat_getrlimit_kvm(procstat->kd, kp, which, rlimit)); case PROCSTAT_SYSCTL: return (procstat_getrlimit_sysctl(kp->ki_pid, which, rlimit)); case PROCSTAT_CORE: return (procstat_getrlimit_core(procstat->core, which, rlimit)); default: warnx("unknown access method: %d", procstat->type); return (-1); } } static int procstat_getpathname_sysctl(pid_t pid, char *pathname, size_t maxlen) { int error, name[4]; size_t len; name[0] = CTL_KERN; name[1] = KERN_PROC; name[2] = KERN_PROC_PATHNAME; name[3] = pid; len = maxlen; error = sysctl(name, 4, pathname, &len, NULL, 0); if (error != 0 && errno != ESRCH) warn("sysctl: kern.proc.pathname: %d", pid); if (len == 0) pathname[0] = '\0'; return (error); } static int procstat_getpathname_core(struct procstat_core *core, char *pathname, size_t maxlen) { struct kinfo_file *files; int cnt, i, result; files = kinfo_getfile_core(core, &cnt); if (files == NULL) return (-1); result = -1; for (i = 0; i < cnt; i++) { if (files[i].kf_fd != KF_FD_TYPE_TEXT) continue; strncpy(pathname, files[i].kf_path, maxlen); result = 0; break; } free(files); return (result); } int procstat_getpathname(struct procstat *procstat, struct kinfo_proc *kp, char *pathname, size_t maxlen) { switch(procstat->type) { case PROCSTAT_KVM: /* XXX: Return empty string. */ if (maxlen > 0) pathname[0] = '\0'; return (0); case PROCSTAT_SYSCTL: return (procstat_getpathname_sysctl(kp->ki_pid, pathname, maxlen)); case PROCSTAT_CORE: return (procstat_getpathname_core(procstat->core, pathname, maxlen)); default: warnx("unknown access method: %d", procstat->type); return (-1); } } static int procstat_getosrel_kvm(kvm_t *kd, struct kinfo_proc *kp, int *osrelp) { struct proc proc; assert(kd != NULL); assert(kp != NULL); if (!kvm_read_all(kd, (unsigned long)kp->ki_paddr, &proc, sizeof(proc))) { warnx("can't read proc struct at %p for pid %d", kp->ki_paddr, kp->ki_pid); return (-1); } *osrelp = proc.p_osrel; return (0); } static int procstat_getosrel_sysctl(pid_t pid, int *osrelp) { int error, name[4]; size_t len; name[0] = CTL_KERN; name[1] = KERN_PROC; name[2] = KERN_PROC_OSREL; name[3] = pid; len = sizeof(*osrelp); error = sysctl(name, 4, osrelp, &len, NULL, 0); if (error != 0 && errno != ESRCH) warn("sysctl: kern.proc.osrel: %d", pid); return (error); } static int procstat_getosrel_core(struct procstat_core *core, int *osrelp) { size_t len; int *buf; buf = procstat_core_get(core, PSC_TYPE_OSREL, NULL, &len); if (buf == NULL) return (-1); if (len < sizeof(*osrelp)) { free(buf); return (-1); } *osrelp = *buf; free(buf); return (0); } int procstat_getosrel(struct procstat *procstat, struct kinfo_proc *kp, int *osrelp) { switch(procstat->type) { case PROCSTAT_KVM: return (procstat_getosrel_kvm(procstat->kd, kp, osrelp)); case PROCSTAT_SYSCTL: return (procstat_getosrel_sysctl(kp->ki_pid, osrelp)); case PROCSTAT_CORE: return (procstat_getosrel_core(procstat->core, osrelp)); default: warnx("unknown access method: %d", procstat->type); return (-1); } } #define PROC_AUXV_MAX 256 #if __ELF_WORD_SIZE == 64 static const char *elf32_sv_names[] = { "Linux ELF32", "FreeBSD ELF32", }; static int is_elf32_sysctl(pid_t pid) { int error, name[4]; size_t len, i; static char sv_name[256]; name[0] = CTL_KERN; name[1] = KERN_PROC; name[2] = KERN_PROC_SV_NAME; name[3] = pid; len = sizeof(sv_name); error = sysctl(name, 4, sv_name, &len, NULL, 0); if (error != 0 || len == 0) return (0); for (i = 0; i < sizeof(elf32_sv_names) / sizeof(*elf32_sv_names); i++) { if (strncmp(sv_name, elf32_sv_names[i], sizeof(sv_name)) == 0) return (1); } return (0); } static Elf_Auxinfo * procstat_getauxv32_sysctl(pid_t pid, unsigned int *cntp) { Elf_Auxinfo *auxv; Elf32_Auxinfo *auxv32; void *ptr; size_t len; unsigned int i, count; int name[4]; name[0] = CTL_KERN; name[1] = KERN_PROC; name[2] = KERN_PROC_AUXV; name[3] = pid; len = PROC_AUXV_MAX * sizeof(Elf32_Auxinfo); auxv = NULL; auxv32 = malloc(len); if (auxv32 == NULL) { warn("malloc(%zu)", len); goto out; } if (sysctl(name, 4, auxv32, &len, NULL, 0) == -1) { if (errno != ESRCH && errno != EPERM) warn("sysctl: kern.proc.auxv: %d: %d", pid, errno); goto out; } count = len / sizeof(Elf_Auxinfo); auxv = malloc(count * sizeof(Elf_Auxinfo)); if (auxv == NULL) { warn("malloc(%zu)", count * sizeof(Elf_Auxinfo)); goto out; } for (i = 0; i < count; i++) { /* * XXX: We expect that values for a_type on a 32-bit platform * are directly mapped to values on 64-bit one, which is not * necessarily true. */ auxv[i].a_type = auxv32[i].a_type; ptr = &auxv32[i].a_un; auxv[i].a_un.a_val = *((uint32_t *)ptr); } *cntp = count; out: free(auxv32); return (auxv); } #endif /* __ELF_WORD_SIZE == 64 */ static Elf_Auxinfo * procstat_getauxv_sysctl(pid_t pid, unsigned int *cntp) { Elf_Auxinfo *auxv; int name[4]; size_t len; #if __ELF_WORD_SIZE == 64 if (is_elf32_sysctl(pid)) return (procstat_getauxv32_sysctl(pid, cntp)); #endif name[0] = CTL_KERN; name[1] = KERN_PROC; name[2] = KERN_PROC_AUXV; name[3] = pid; len = PROC_AUXV_MAX * sizeof(Elf_Auxinfo); auxv = malloc(len); if (auxv == NULL) { warn("malloc(%zu)", len); return (NULL); } if (sysctl(name, 4, auxv, &len, NULL, 0) == -1) { if (errno != ESRCH && errno != EPERM) warn("sysctl: kern.proc.auxv: %d: %d", pid, errno); free(auxv); return (NULL); } *cntp = len / sizeof(Elf_Auxinfo); return (auxv); } static Elf_Auxinfo * procstat_getauxv_core(struct procstat_core *core, unsigned int *cntp) { Elf_Auxinfo *auxv; size_t len; auxv = procstat_core_get(core, PSC_TYPE_AUXV, NULL, &len); if (auxv == NULL) return (NULL); *cntp = len / sizeof(Elf_Auxinfo); return (auxv); } Elf_Auxinfo * procstat_getauxv(struct procstat *procstat, struct kinfo_proc *kp, unsigned int *cntp) { switch(procstat->type) { case PROCSTAT_KVM: warnx("kvm method is not supported"); return (NULL); case PROCSTAT_SYSCTL: return (procstat_getauxv_sysctl(kp->ki_pid, cntp)); case PROCSTAT_CORE: return (procstat_getauxv_core(procstat->core, cntp)); default: warnx("unknown access method: %d", procstat->type); return (NULL); } } void procstat_freeauxv(struct procstat *procstat __unused, Elf_Auxinfo *auxv) { free(auxv); } static struct kinfo_kstack * procstat_getkstack_sysctl(pid_t pid, int *cntp) { struct kinfo_kstack *kkstp; int error, name[4]; size_t len; name[0] = CTL_KERN; name[1] = KERN_PROC; name[2] = KERN_PROC_KSTACK; name[3] = pid; len = 0; error = sysctl(name, 4, NULL, &len, NULL, 0); if (error < 0 && errno != ESRCH && errno != EPERM && errno != ENOENT) { warn("sysctl: kern.proc.kstack: %d", pid); return (NULL); } if (error == -1 && errno == ENOENT) { warnx("sysctl: kern.proc.kstack unavailable" " (options DDB or options STACK required in kernel)"); return (NULL); } if (error == -1) return (NULL); kkstp = malloc(len); if (kkstp == NULL) { warn("malloc(%zu)", len); return (NULL); } if (sysctl(name, 4, kkstp, &len, NULL, 0) == -1) { warn("sysctl: kern.proc.pid: %d", pid); free(kkstp); return (NULL); } *cntp = len / sizeof(*kkstp); return (kkstp); } struct kinfo_kstack * procstat_getkstack(struct procstat *procstat, struct kinfo_proc *kp, unsigned int *cntp) { switch(procstat->type) { case PROCSTAT_KVM: warnx("kvm method is not supported"); return (NULL); case PROCSTAT_SYSCTL: return (procstat_getkstack_sysctl(kp->ki_pid, cntp)); case PROCSTAT_CORE: warnx("core method is not supported"); return (NULL); default: warnx("unknown access method: %d", procstat->type); return (NULL); } } void procstat_freekstack(struct procstat *procstat __unused, struct kinfo_kstack *kkstp) { free(kkstp); } Index: projects/clang-trunk/libexec/ypxfr/ypxfr_main.c =================================================================== --- projects/clang-trunk/libexec/ypxfr/ypxfr_main.c (revision 283870) +++ projects/clang-trunk/libexec/ypxfr/ypxfr_main.c (revision 283871) @@ -1,577 +1,577 @@ /* * Copyright (c) 1995 * Bill Paul . All rights reserved. * * Redistribution and use in source and binary forms, with or without * modification, are permitted provided that the following conditions * are met: * 1. Redistributions of source code must retain the above copyright * notice, this list of conditions and the following disclaimer. * 2. Redistributions in binary form must reproduce the above copyright * notice, this list of conditions and the following disclaimer in the * documentation and/or other materials provided with the distribution. * 3. All advertising materials mentioning features or use of this software * must display the following acknowledgement: * This product includes software developed by Bill Paul. * 4. Neither the name of the author nor the names of any co-contributors * may be used to endorse or promote products derived from this software * without specific prior written permission. * * THIS SOFTWARE IS PROVIDED BY Bill Paul AND CONTRIBUTORS ``AS IS'' AND * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE * ARE DISCLAIMED. IN NO EVENT SHALL Bill Paul OR CONTRIBUTORS BE LIABLE * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF * SUCH DAMAGE. */ #include __FBSDID("$FreeBSD$"); #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include "ypxfr_extern.h" char *progname = "ypxfr"; char *yp_dir = _PATH_YP; int _rpcpmstart = 0; int ypxfr_use_yplib = 0; /* Assume the worst. */ int ypxfr_clear = 1; int ypxfr_prognum = 0; struct sockaddr_in ypxfr_callback_addr; struct yppushresp_xfr ypxfr_resp; DB *dbp; static void ypxfr_exit(ypxfrstat retval, char *temp) { CLIENT *clnt; int sock = RPC_ANYSOCK; struct timeval timeout; /* Clean up no matter what happened previously. */ if (temp != NULL) { if (dbp != NULL) (void)(dbp->close)(dbp); if (unlink(temp) == -1) { yp_error("failed to unlink %s",strerror(errno)); } } if (ypxfr_prognum) { timeout.tv_sec = 20; timeout.tv_usec = 0; if ((clnt = clntudp_create(&ypxfr_callback_addr, ypxfr_prognum, 1, timeout, &sock)) == NULL) { yp_error("%s", clnt_spcreateerror("failed to " "establish callback handle")); exit(1); } ypxfr_resp.status = (yppush_status)retval; if (yppushproc_xfrresp_1(&ypxfr_resp, clnt) == NULL) { yp_error("%s", clnt_sperror(clnt, "callback failed")); clnt_destroy(clnt); exit(1); } clnt_destroy(clnt); } else { yp_error("Exiting: %s", ypxfrerr_string(retval)); } exit(0); } static void usage(void) { if (_rpcpmstart) { ypxfr_exit(YPXFR_BADARGS,NULL); } else { fprintf(stderr, "%s\n%s\n%s\n", "usage: ypxfr [-f] [-c] [-d target domain] [-h source host]", " [-s source domain] [-p path]", " [-C taskid program-number ipaddr port] mapname"); exit(1); } } int ypxfr_foreach(int status, char *key, int keylen, char *val, int vallen, char *data) { DBT dbkey, dbval; if (status != YP_TRUE) return (status); /* * XXX Do not attempt to write zero-length keys or * data into a Berkeley DB hash database. It causes a * strange failure mode where sequential searches get * caught in an infinite loop. */ if (keylen) { dbkey.data = key; dbkey.size = keylen; } else { dbkey.data = ""; dbkey.size = 1; } if (vallen) { dbval.data = val; dbval.size = vallen; } else { dbval.data = ""; dbval.size = 1; } if (yp_put_record(dbp, &dbkey, &dbval, 0) != YP_TRUE) return(yp_errno); return (0); } int main(int argc, char *argv[]) { int ch; int ypxfr_force = 0; char *ypxfr_dest_domain = NULL; char *ypxfr_source_host = NULL; char *ypxfr_source_domain = NULL; char *ypxfr_local_domain = NULL; char *ypxfr_master = NULL; unsigned long ypxfr_order = -1, ypxfr_skew_check = -1; char *ypxfr_mapname = NULL; int ypxfr_args = 0; char ypxfr_temp_map[MAXPATHLEN + 2]; char tempmap[MAXPATHLEN + 2]; char buf[MAXPATHLEN + 2]; DBT key, data; int remoteport; int interdom = 0; int secure = 0; debug = 1; if (!isatty(fileno(stderr))) { openlog("ypxfr", LOG_PID, LOG_DAEMON); _rpcpmstart = 1; } if (argc < 2) usage(); while ((ch = getopt(argc, argv, "fcd:h:s:p:C:")) != -1) { int my_optind; switch (ch) { case 'f': ypxfr_force++; ypxfr_args++; break; case 'c': ypxfr_clear = 0; ypxfr_args++; break; case 'd': ypxfr_dest_domain = optarg; ypxfr_args += 2; break; case 'h': ypxfr_source_host = optarg; ypxfr_args += 2; break; case 's': ypxfr_source_domain = optarg; ypxfr_args += 2; break; case 'p': yp_dir = optarg; ypxfr_args += 2; break; case 'C': /* * Whoever decided that the -C flag should take * four arguments is a twit. */ my_optind = optind - 1; if (argv[my_optind] == NULL || !strlen(argv[my_optind])) { yp_error("transaction ID not specified"); usage(); } ypxfr_resp.transid = atol(argv[my_optind]); my_optind++; if (argv[my_optind] == NULL || !strlen(argv[my_optind])) { yp_error("RPC program number not specified"); usage(); } ypxfr_prognum = atol(argv[my_optind]); my_optind++; if (argv[my_optind] == NULL || !strlen(argv[my_optind])) { yp_error("address not specified"); usage(); } if (!inet_aton(argv[my_optind], &ypxfr_callback_addr.sin_addr)) { yp_error("failed to convert '%s' to IP addr", argv[my_optind]); exit(1); } my_optind++; if (argv[my_optind] == NULL || !strlen(argv[my_optind])) { yp_error("port not specified"); usage(); } ypxfr_callback_addr.sin_port = htons((u_short)atoi(argv[my_optind])); ypxfr_args += 5; break; default: usage(); break; } } ypxfr_mapname = argv[ypxfr_args + 1]; if (ypxfr_mapname == NULL) { yp_error("no map name specified"); usage(); } /* Always the case. */ ypxfr_callback_addr.sin_family = AF_INET; /* Determine if local NIS client facilities are turned on. */ if (!yp_get_default_domain(&ypxfr_local_domain) && _yp_check(&ypxfr_local_domain)) ypxfr_use_yplib = 1; /* * If no destination domain is specified, assume that the * local default domain is to be used and try to obtain it. * Fails if NIS client facilities are turned off. */ if (ypxfr_dest_domain == NULL) { if (ypxfr_use_yplib) { yp_get_default_domain(&ypxfr_dest_domain); } else { yp_error("no destination domain specified and \ the local domain name isn't set"); ypxfr_exit(YPXFR_BADARGS,NULL); } } /* * If a source domain is not specified, assume it to * be the same as the destination domain. */ if (ypxfr_source_domain == NULL) { ypxfr_source_domain = ypxfr_dest_domain; } /* * If the source host is not specified, assume it to be the * master for the specified map. If local NIS client facilities * are turned on, we can figure this out using yp_master(). * If not, we have to see if a local copy of the map exists * and extract its YP_MASTER_NAME record. If _that_ fails, * we are stuck and must ask the user for more information. */ if (ypxfr_source_host == NULL) { if (!ypxfr_use_yplib) { /* * Double whammy: NIS isn't turned on and the user * didn't specify a source host. */ char *dptr; key.data = "YP_MASTER_NAME"; key.size = sizeof("YP_MASTER_NAME") - 1; if (yp_get_record(ypxfr_dest_domain, ypxfr_mapname, &key, &data, 1) != YP_TRUE) { yp_error("no source host specified"); ypxfr_exit(YPXFR_BADARGS,NULL); } dptr = data.data; dptr[data.size] = '\0'; ypxfr_master = ypxfr_source_host = strdup(dptr); } } else { if (ypxfr_use_yplib) ypxfr_use_yplib = 0; } if (ypxfr_master == NULL) { if ((ypxfr_master = ypxfr_get_master(ypxfr_source_domain, ypxfr_mapname, ypxfr_source_host, ypxfr_use_yplib)) == NULL) { yp_error("failed to find master of %s in domain %s: %s", ypxfr_mapname, ypxfr_source_domain, ypxfrerr_string((ypxfrstat)yp_errno)); ypxfr_exit(YPXFR_MADDR,NULL); } } /* * If we got here and ypxfr_source_host is still undefined, * it means we had to resort to using yp_master() to find the * master server for the map. The source host and master should * be identical. */ if (ypxfr_source_host == NULL) ypxfr_source_host = ypxfr_master; /* * Don't talk to ypservs on unprivileged ports. */ remoteport = getrpcport(ypxfr_source_host, YPPROG, YPVERS, IPPROTO_UDP); if (remoteport >= IPPORT_RESERVED) { yp_error("ypserv on %s not running on reserved port", ypxfr_source_host); ypxfr_exit(YPXFR_REFUSED, NULL); } if ((ypxfr_order = ypxfr_get_order(ypxfr_source_domain, ypxfr_mapname, ypxfr_master, 0)) == 0) { yp_error("failed to get order number of %s: %s", - ypxfr_mapname, yp_errno == YPXFR_SUCC ? + ypxfr_mapname, yp_errno == YP_TRUE ? "map has order 0" : ypxfrerr_string((ypxfrstat)yp_errno)); ypxfr_exit(YPXFR_YPERR,NULL); } if (ypxfr_match(ypxfr_master, ypxfr_source_domain, ypxfr_mapname, "YP_INTERDOMAIN", sizeof("YP_INTERDOMAIN") - 1)) interdom++; if (ypxfr_match(ypxfr_master, ypxfr_source_domain, ypxfr_mapname, "YP_SECURE", sizeof("YP_SECURE") - 1)) secure++; key.data = "YP_LAST_MODIFIED"; key.size = sizeof("YP_LAST_MODIFIED") - 1; /* The order number is immaterial when the 'force' flag is set. */ if (!ypxfr_force) { int ignore = 0; if (yp_get_record(ypxfr_dest_domain,ypxfr_mapname,&key,&data,1) != YP_TRUE) { switch (yp_errno) { case YP_NOKEY: ypxfr_exit(YPXFR_FORCE,NULL); break; case YP_NOMAP: /* * If the map doesn't exist, we're * creating it. Ignore the error. */ ignore++; break; case YP_BADDB: default: ypxfr_exit(YPXFR_DBM,NULL); break; } } if (!ignore && ypxfr_order <= atoi(data.data)) ypxfr_exit(YPXFR_AGE, NULL); } /* Construct a temporary map file name */ snprintf(tempmap, sizeof(tempmap), "%s.%d",ypxfr_mapname, getpid()); snprintf(ypxfr_temp_map, sizeof(ypxfr_temp_map), "%s/%s/%s", yp_dir, ypxfr_dest_domain, tempmap); if ((remoteport = getrpcport(ypxfr_source_host, YPXFRD_FREEBSD_PROG, YPXFRD_FREEBSD_VERS, IPPROTO_TCP))) { /* Don't talk to rpc.ypxfrds on unprovileged ports. */ if (remoteport >= IPPORT_RESERVED) { yp_error("rpc.ypxfrd on %s not using privileged port", ypxfr_source_host); ypxfr_exit(YPXFR_REFUSED, NULL); } /* Try to send using ypxfrd. If it fails, use old method. */ if (!ypxfrd_get_map(ypxfr_source_host, ypxfr_mapname, ypxfr_source_domain, ypxfr_temp_map)) goto leave; } /* Open the temporary map read/write. */ if ((dbp = yp_open_db_rw(ypxfr_dest_domain, tempmap, 0)) == NULL) { yp_error("failed to open temporary map file"); ypxfr_exit(YPXFR_DBM,NULL); } /* * Fill in the keys we already know, such as the order number, * master name, input file name (we actually make up a bogus * name for that) and output file name. */ snprintf(buf, sizeof(buf), "%lu", ypxfr_order); data.data = buf; data.size = strlen(buf); if (yp_put_record(dbp, &key, &data, 0) != YP_TRUE) { yp_error("failed to write order number to database"); ypxfr_exit(YPXFR_DBM,ypxfr_temp_map); } key.data = "YP_MASTER_NAME"; key.size = sizeof("YP_MASTER_NAME") - 1; data.data = ypxfr_master; data.size = strlen(ypxfr_master); if (yp_put_record(dbp, &key, &data, 0) != YP_TRUE) { yp_error("failed to write master name to database"); ypxfr_exit(YPXFR_DBM,ypxfr_temp_map); } key.data = "YP_DOMAIN_NAME"; key.size = sizeof("YP_DOMAIN_NAME") - 1; data.data = ypxfr_dest_domain; data.size = strlen(ypxfr_dest_domain); if (yp_put_record(dbp, &key, &data, 0) != YP_TRUE) { yp_error("failed to write domain name to database"); ypxfr_exit(YPXFR_DBM,ypxfr_temp_map); } snprintf (buf, sizeof(buf), "%s:%s", ypxfr_source_host, ypxfr_mapname); key.data = "YP_INPUT_NAME"; key.size = sizeof("YP_INPUT_NAME") - 1; data.data = &buf; data.size = strlen(buf); if (yp_put_record(dbp, &key, &data, 0) != YP_TRUE) { yp_error("failed to write input name to database"); ypxfr_exit(YPXFR_DBM,ypxfr_temp_map); } snprintf(buf, sizeof(buf), "%s/%s/%s", yp_dir, ypxfr_dest_domain, ypxfr_mapname); key.data = "YP_OUTPUT_NAME"; key.size = sizeof("YP_OUTPUT_NAME") - 1; data.data = &buf; data.size = strlen(buf); if (yp_put_record(dbp, &key, &data, 0) != YP_TRUE) { yp_error("failed to write output name to database"); ypxfr_exit(YPXFR_DBM,ypxfr_temp_map); } if (interdom) { key.data = "YP_INTERDOMAIN"; key.size = sizeof("YP_INTERDOMAIN") - 1; data.data = ""; data.size = 0; if (yp_put_record(dbp, &key, &data, 0) != YP_TRUE) { yp_error("failed to add interdomain flag to database"); ypxfr_exit(YPXFR_DBM,ypxfr_temp_map); } } if (secure) { key.data = "YP_SECURE"; key.size = sizeof("YP_SECURE") - 1; data.data = ""; data.size = 0; if (yp_put_record(dbp, &key, &data, 0) != YP_TRUE) { yp_error("failed to add secure flag to database"); ypxfr_exit(YPXFR_DBM,ypxfr_temp_map); } } /* Now suck over the contents of the map from the master. */ if (ypxfr_get_map(ypxfr_mapname,ypxfr_source_domain, ypxfr_source_host, ypxfr_foreach)){ yp_error("failed to retrieve map from source host"); ypxfr_exit(YPXFR_YPERR,ypxfr_temp_map); } (void)(dbp->close)(dbp); dbp = NULL; /* <- yes, it seems this is necessary. */ leave: snprintf(buf, sizeof(buf), "%s/%s/%s", yp_dir, ypxfr_dest_domain, ypxfr_mapname); /* Peek at the order number again and check for skew. */ if ((ypxfr_skew_check = ypxfr_get_order(ypxfr_source_domain, ypxfr_mapname, ypxfr_master, 0)) == 0) { yp_error("failed to get order number of %s: %s", - ypxfr_mapname, yp_errno == YPXFR_SUCC ? + ypxfr_mapname, yp_errno == YP_TRUE ? "map has order 0" : ypxfrerr_string((ypxfrstat)yp_errno)); ypxfr_exit(YPXFR_YPERR,ypxfr_temp_map); } if (ypxfr_order != ypxfr_skew_check) ypxfr_exit(YPXFR_SKEW,ypxfr_temp_map); /* * Send a YPPROC_CLEAR to the local ypserv. */ if (ypxfr_clear) { char in = 0; char *out = NULL; int stat; if ((stat = callrpc("localhost",YPPROG,YPVERS,YPPROC_CLEAR, (xdrproc_t)xdr_void, (void *)&in, (xdrproc_t)xdr_void, (void *)out)) != RPC_SUCCESS) { yp_error("failed to send 'clear' to local ypserv: %s", clnt_sperrno((enum clnt_stat) stat)); ypxfr_exit(YPXFR_CLEAR, ypxfr_temp_map); } } /* * Put the new map in place immediately. I'm not sure if the * kernel does an unlink() and rename() atomically in the event * that we move a new copy of a map over the top of an existing * one, but there's less chance of a race condition happening * than if we were to do the unlink() ourselves. */ if (rename(ypxfr_temp_map, buf) == -1) { yp_error("rename(%s,%s) failed: %s", ypxfr_temp_map, buf, strerror(errno)); ypxfr_exit(YPXFR_FILE,NULL); } ypxfr_exit(YPXFR_SUCC,NULL); return(1); } Index: projects/clang-trunk/share/man/man5/src.conf.5 =================================================================== --- projects/clang-trunk/share/man/man5/src.conf.5 (revision 283870) +++ projects/clang-trunk/share/man/man5/src.conf.5 (revision 283871) @@ -1,1281 +1,1288 @@ .\" DO NOT EDIT-- this file is automatically generated. .\" from FreeBSD: head/tools/build/options/makeman 255964 2013-10-01 07:22:04Z des .\" $FreeBSD$ -.Dd April 19, 2015 +.Dd May 30, 2015 .Dt SRC.CONF 5 .Os .Sh NAME .Nm src.conf .Nd "source build options" .Sh DESCRIPTION The .Nm file contains settings that will apply to every build involving the .Fx source tree; see .Xr build 7 . .Pp The .Nm file uses the standard makefile syntax. However, .Nm should not specify any dependencies to .Xr make 1 . Instead, .Nm is to set .Xr make 1 variables that control the aspects of how the system builds. .Pp The default location of .Nm is .Pa /etc/src.conf , though an alternative location can be specified in the .Xr make 1 variable .Va SRCCONF . Overriding the location of .Nm may be necessary if the system-wide settings are not suitable for a particular build. For instance, setting .Va SRCCONF to .Pa /dev/null effectively resets all build controls to their defaults. .Pp The only purpose of .Nm is to control the compilation of the .Fx source code, which is usually located in .Pa /usr/src . As a rule, the system administrator creates .Nm when the values of certain control variables need to be changed from their defaults. .Pp In addition, control variables can be specified for a particular build via the .Fl D option of .Xr make 1 or in its environment; see .Xr environ 7 . .Pp The values of variables are ignored regardless of their setting; even if they would be set to .Dq Li FALSE or .Dq Li NO . Just the existence of an option will cause it to be honoured by .Xr make 1 . .Pp The following list provides a name and short description for variables that can be used for source builds. .Bl -tag -width indent .It Va WITHOUT_ACCT .\" from FreeBSD: head/tools/build/options/WITHOUT_ACCT 223201 2011-06-17 20:47:44Z ed Set to not build process accounting tools such as .Xr accton 8 and .Xr sa 8 . .It Va WITHOUT_ACPI .\" from FreeBSD: head/tools/build/options/WITHOUT_ACPI 156932 2006-03-21 07:50:50Z ru Set to not build .Xr acpiconf 8 , .Xr acpidump 8 and related programs. .It Va WITHOUT_AMD .\" from FreeBSD: head/tools/build/options/WITHOUT_AMD 183242 2008-09-21 22:02:26Z sam Set to not build .Xr amd 8 , and related programs. .It Va WITHOUT_APM .\" from FreeBSD: head/tools/build/options/WITHOUT_APM 183242 2008-09-21 22:02:26Z sam Set to not build .Xr apm 8 , .Xr apmd 8 and related programs. .It Va WITHOUT_ASSERT_DEBUG .\" from FreeBSD: head/tools/build/options/WITHOUT_ASSERT_DEBUG 162215 2006-09-11 13:55:27Z ru Set to compile programs and libraries without the .Xr assert 3 checks. .It Va WITHOUT_AT .\" from FreeBSD: head/tools/build/options/WITHOUT_AT 183242 2008-09-21 22:02:26Z sam Set to not build .Xr at 1 and related utilities. .It Va WITHOUT_ATM .\" from FreeBSD: head/tools/build/options/WITHOUT_ATM 156932 2006-03-21 07:50:50Z ru Set to not build programs and libraries related to ATM networking. .It Va WITHOUT_AUDIT .\" from FreeBSD: head/tools/build/options/WITHOUT_AUDIT 156932 2006-03-21 07:50:50Z ru Set to not build audit support into system programs. .It Va WITHOUT_AUTHPF .\" from FreeBSD: head/tools/build/options/WITHOUT_AUTHPF 156932 2006-03-21 07:50:50Z ru Set to not build .Xr authpf 8 . .It Va WITHOUT_AUTOFS .\" from FreeBSD: head/tools/build/options/WITHOUT_AUTOFS 277728 2015-01-26 07:15:49Z ngie Set to not build .Xr autofs 4 related programs, libraries, and kernel modules. .It Va WITHOUT_BHYVE .\" from FreeBSD: head/tools/build/options/WITHOUT_BHYVE 277727 2015-01-26 06:44:48Z ngie Set to not build or install .Xr bhyve 8 , associated utilities, and examples. .Pp This option only affects amd64/amd64. .It Va WITHOUT_BINUTILS .\" from FreeBSD: head/tools/build/options/WITHOUT_BINUTILS 266158 2014-05-15 16:51:45Z brooks Set to not build or install binutils (as, c++-filt, gconv, ld, nm, objcopy, objdump, readelf, size and strip) as part of the normal system build. The resulting system cannot build programs from source. .It Va WITHOUT_BINUTILS_BOOTSTRAP .\" from FreeBSD: head/tools/build/options/WITHOUT_BINUTILS_BOOTSTRAP 264660 2014-04-18 17:03:58Z imp Set to not build binutils (as, c++-filt, gconv, ld, nm, objcopy, objdump, readelf, size and strip) as part of the bootstrap process. .Bf -symbolic The option does not work for build targets unless some alternative toolchain is provided. .Ef .It Va WITHOUT_BLUETOOTH .\" from FreeBSD: head/tools/build/options/WITHOUT_BLUETOOTH 156932 2006-03-21 07:50:50Z ru Set to not build Bluetooth related kernel modules, programs and libraries. .It Va WITHOUT_BOOT .\" from FreeBSD: head/tools/build/options/WITHOUT_BOOT 156932 2006-03-21 07:50:50Z ru Set to not build the boot blocks and loader. .It Va WITHOUT_BOOTPARAMD .\" from FreeBSD: head/tools/build/options/WITHOUT_BOOTPARAMD 278192 2015-02-04 10:19:32Z ngie Set to not build or install .Xr bootparamd 8 . .It Va WITHOUT_BOOTPD .\" from FreeBSD: head/tools/build/options/WITHOUT_BOOTPD 278192 2015-02-04 10:19:32Z ngie Set to not build or install .Xr bootpd 8 . .It Va WITHOUT_BSDINSTALL .\" from FreeBSD: head/tools/build/options/WITHOUT_BSDINSTALL 277677 2015-01-25 04:43:13Z ngie Set to not build .Xr bsdinstall 8 , .Xr sade 8 , and related programs. .It Va WITHOUT_BSD_CPIO .\" from FreeBSD: head/tools/build/options/WITHOUT_BSD_CPIO 179813 2008-06-16 05:48:15Z dougb Set to not build the BSD licensed version of cpio based on .Xr libarchive 3 . .It Va WITH_BSD_GREP .\" from FreeBSD: head/tools/build/options/WITH_BSD_GREP 222273 2011-05-25 01:04:12Z obrien Install BSD-licensed grep as '[ef]grep' instead of GNU grep. .It Va WITHOUT_BSNMP .\" from FreeBSD: head/tools/build/options/WITHOUT_BSNMP 183306 2008-09-23 16:15:42Z sam Set to not build or install .Xr bsnmpd 1 and related libraries and data files. .It Va WITHOUT_BZIP2 .\" from FreeBSD: head/tools/build/options/WITHOUT_BZIP2 174550 2007-12-12 16:43:17Z ru Set to not build contributed bzip2 software as a part of the base system. .Bf -symbolic The option has no effect yet. .Ef When set, it also enforces the following options: .Pp .Bl -item -compact .It .Va WITHOUT_BZIP2_SUPPORT .El .It Va WITHOUT_BZIP2_SUPPORT .\" from FreeBSD: head/tools/build/options/WITHOUT_BZIP2_SUPPORT 166255 2007-01-26 10:19:08Z delphij Set to build some programs without optional bzip2 support. .It Va WITHOUT_CALENDAR .\" from FreeBSD: head/tools/build/options/WITHOUT_CALENDAR 156932 2006-03-21 07:50:50Z ru Set to not build .Xr calendar 1 . .It Va WITHOUT_CAPSICUM .\" from FreeBSD: head/tools/build/options/WITHOUT_CAPSICUM 229319 2012-01-02 21:57:58Z rwatson Set to not build Capsicum support into system programs. .It Va WITHOUT_CASPER .\" from FreeBSD: head/tools/build/options/WITHOUT_CASPER 258838 2013-12-02 08:21:28Z pjd Set to not build Casper program and related libraries. .It Va WITHOUT_CCD .\" from FreeBSD: head/tools/build/options/WITHOUT_CCD 277678 2015-01-25 04:52:48Z ngie Set to not build .Xr geom_ccd 4 and related utilities. .It Va WITHOUT_CDDL .\" from FreeBSD: head/tools/build/options/WITHOUT_CDDL 163861 2006-11-01 09:02:11Z jb Set to not build code licensed under Sun's CDDL. When set, it also enforces the following options: .Pp .Bl -item -compact .It .Va WITHOUT_CTF .It .Va WITHOUT_ZFS .El .It Va WITHOUT_CLANG .\" from FreeBSD: head/tools/build/options/WITHOUT_CLANG 264660 2014-04-18 17:03:58Z imp Set to not build the Clang C/C++ compiler during the regular phase of the build. .Pp It is a default setting on arm/armeb, mips/mipsel, mips/mips, mips/mips64el, mips/mips64, mips/mipsn32 and sparc64/sparc64. When set, it also enforces the following options: .Pp .Bl -item -compact .It .Va WITHOUT_CLANG_EXTRAS .It .Va WITHOUT_CLANG_FULL .El .It Va WITH_CLANG .\" from FreeBSD: head/tools/build/options/WITH_CLANG 264660 2014-04-18 17:03:58Z imp Set to build the Clang C/C++ compiler during the normal phase of the build. .Pp It is a default setting on amd64/amd64, arm/arm, arm/armv6, arm/armv6hf, i386/i386, pc98/i386, powerpc/powerpc and powerpc/powerpc64. .It Va WITHOUT_CLANG_BOOTSTRAP .\" from FreeBSD: head/tools/build/options/WITHOUT_CLANG_BOOTSTRAP 273177 2014-10-16 18:28:11Z skreuzer Set to not build the Clang C/C++ compiler during the bootstrap phase of the build. You must enable either gcc or clang bootstrap to be able to build the system, unless an alternative compiler is provided via XCC. .Pp It is a default setting on arm/armeb, mips/mipsel, mips/mips, mips/mips64el, mips/mips64, mips/mipsn32, powerpc/powerpc, powerpc/powerpc64 and sparc64/sparc64. .It Va WITH_CLANG_BOOTSTRAP .\" from FreeBSD: head/tools/build/options/WITH_CLANG_BOOTSTRAP 264660 2014-04-18 17:03:58Z imp Set to build the Clang C/C++ compiler during the bootstrap phase of the build. .Pp It is a default setting on amd64/amd64, arm/arm, arm/armv6, arm/armv6hf, i386/i386 and pc98/i386. .It Va WITH_CLANG_EXTRAS .\" from FreeBSD: head/tools/build/options/WITH_CLANG_EXTRAS 231057 2012-02-05 23:56:22Z dim Set to build additional clang and llvm tools, such as bugpoint. .It Va WITHOUT_CLANG_FULL .\" from FreeBSD: head/tools/build/options/WITHOUT_CLANG_FULL 246259 2013-02-02 22:28:29Z dim Set to avoid building the ARCMigrate, Rewriter and StaticAnalyzer components of the Clang C/C++ compiler. .Pp It is a default setting on arm/arm, arm/armeb, arm/armv6, arm/armv6hf, mips/mipsel, mips/mips, mips/mips64el, mips/mips64, mips/mipsn32 and sparc64/sparc64. .It Va WITH_CLANG_FULL .\" from FreeBSD: head/tools/build/options/WITH_CLANG_FULL 246259 2013-02-02 22:28:29Z dim Set to build the ARCMigrate, Rewriter and StaticAnalyzer components of the Clang C/C++ compiler. .Pp It is a default setting on amd64/amd64, i386/i386, pc98/i386, powerpc/powerpc and powerpc/powerpc64. .It Va WITHOUT_CLANG_IS_CC .\" from FreeBSD: head/tools/build/options/WITHOUT_CLANG_IS_CC 242629 2012-11-05 21:53:23Z brooks Set to install the GCC compiler as .Pa /usr/bin/cc , .Pa /usr/bin/c++ and .Pa /usr/bin/cpp . .Pp It is a default setting on arm/armeb, mips/mipsel, mips/mips, mips/mips64el, mips/mips64, mips/mipsn32, powerpc/powerpc, powerpc/powerpc64 and sparc64/sparc64. .It Va WITH_CLANG_IS_CC .\" from FreeBSD: head/tools/build/options/WITH_CLANG_IS_CC 235342 2012-05-12 16:12:36Z gjb Set to install the Clang C/C++ compiler as .Pa /usr/bin/cc , .Pa /usr/bin/c++ and .Pa /usr/bin/cpp . .Pp It is a default setting on amd64/amd64, arm/arm, arm/armv6, arm/armv6hf, i386/i386 and pc98/i386. .It Va WITHOUT_CPP .\" from FreeBSD: head/tools/build/options/WITHOUT_CPP 156932 2006-03-21 07:50:50Z ru Set to not build .Xr cpp 1 . .It Va WITHOUT_CROSS_COMPILER .\" from FreeBSD: head/tools/build/options/WITHOUT_CROSS_COMPILER 275138 2014-11-26 20:43:09Z gjb Set to not build any cross compiler in the cross-tools stage of buildworld. If you are compiling a different version of .Fx than what is installed on the system, you will need to provide an alternate compiler with XCC to ensure success. If you are compiling with an identical version of .Fx to the host, this option may be safely used. This option may also be safe when the host version of .Fx is close to the sources being built, but all bets are off if there have been any changes to the toolchain between the versions. When set, it also enforces the following options: .Pp .Bl -item -compact .It .Va WITHOUT_BINUTILS_BOOTSTRAP .It .Va WITHOUT_CLANG_BOOTSTRAP .It .Va WITHOUT_GCC_BOOTSTRAP .El .It Va WITHOUT_CRYPT .\" from FreeBSD: head/tools/build/options/WITHOUT_CRYPT 156932 2006-03-21 07:50:50Z ru Set to not build any crypto code. When set, it also enforces the following options: .Pp .Bl -item -compact .It .Va WITHOUT_KERBEROS .It .Va WITHOUT_KERBEROS_SUPPORT .It .Va WITHOUT_OPENSSH .It .Va WITHOUT_OPENSSL .El .Pp When set, the following options are also in effect: .Pp .Bl -inset -compact .It Va WITHOUT_GSSAPI (unless .Va WITH_GSSAPI is set explicitly) .El .It Va WITH_CTF .\" from FreeBSD: head/tools/build/options/WITH_CTF 228159 2011-11-30 18:22:44Z fjoe Set to compile with CTF (Compact C Type Format) data. CTF data encapsulates a reduced form of debugging information similar to DWARF and the venerable stabs and is required for DTrace. .It Va WITHOUT_CTM .\" from FreeBSD: head/tools/build/options/WITHOUT_CTM 183242 2008-09-21 22:02:26Z sam Set to not build .Xr ctm 1 and related utilities. .It Va WITHOUT_CUSE .\" from FreeBSD: head/tools/build/options/WITHOUT_CUSE 270171 2014-08-19 15:40:26Z hselasky Set to not build CUSE-related programs and libraries. .It Va WITHOUT_CXX .\" from FreeBSD: head/tools/build/options/WITHOUT_CXX 281053 2015-04-03 23:55:04Z bdrewery Set to not build .Xr c++ 1 and related libraries. It will also prevent building of .Xr gperf 1 and .Xr devd 8 . When set, it also enforces the following options: .Pp .Bl -item -compact .It .Va WITHOUT_CLANG .It .Va WITHOUT_CLANG_EXTRAS .It .Va WITHOUT_CLANG_FULL .It .Va WITHOUT_GNUCXX .It .Va WITHOUT_GROFF .El .It Va WITH_DEBUG_FILES .\" from FreeBSD: head/tools/build/options/WITH_DEBUG_FILES 251512 2013-06-07 21:40:02Z emaste Set to strip debug info into a separate file for each executable binary and shared library. The debug files will be placed in a subdirectory of .Pa /usr/lib/debug and are located automatically by .Xr gdb 1 . .It Va WITHOUT_DICT .\" from FreeBSD: head/tools/build/options/WITHOUT_DICT 156932 2006-03-21 07:50:50Z ru Set to not build the Webster dictionary files. .It Va WITHOUT_DMAGENT .\" from FreeBSD: head/tools/build/options/WITHOUT_DMAGENT 262335 2014-02-22 13:05:23Z bapt Set to not build dma Mail Transport Agent .It Va WITHOUT_DOCCOMPRESS .\" from FreeBSD: head/tools/build/options/WITHOUT_DOCCOMPRESS 266752 2014-05-27 15:52:27Z gjb Set to not to install compressed system documentation. Only the uncompressed version will be installed. .It Va WITHOUT_DYNAMICROOT .\" from FreeBSD: head/tools/build/options/WITHOUT_DYNAMICROOT 156932 2006-03-21 07:50:50Z ru Set this if you do not want to link .Pa /bin and .Pa /sbin dynamically. .It Va WITHOUT_ED_CRYPTO .\" from FreeBSD: head/tools/build/options/WITHOUT_ED_CRYPTO 235660 2012-05-19 20:05:27Z marcel Set to build .Xr ed 1 without support for encryption/decryption. .It Va WITHOUT_EE .\" from FreeBSD: head/tools/build/options/WITHOUT_EE 277663 2015-01-25 00:03:44Z ngie Set to not build and install .Xr edit 1 , .Xr ee 1 , and related programs. .It Va WITH_EISA .\" from FreeBSD: head/tools/build/options/WITH_EISA 264654 2014-04-18 16:53:06Z imp Set to build EISA kernel modules. .It Va WITHOUT_ELFTOOLCHAIN_TOOLS .\" from FreeBSD: head/tools/build/options/WITHOUT_ELFTOOLCHAIN_TOOLS 276796 2015-01-07 22:02:37Z emaste Set to use .Xr addr2line 1 , .Xr nm 1 , .Xr size 1 , .Xr strings 1 , and .Xr strip 1 from GNU binutils instead of the ELF Tool Chain project. .It Va WITHOUT_EXAMPLES .\" from FreeBSD: head/tools/build/options/WITHOUT_EXAMPLES 156938 2006-03-21 09:06:24Z ru Set to avoid installing examples to .Pa /usr/share/examples/ . .It Va WITHOUT_FDT .\" from FreeBSD: head/tools/build/options/WITHOUT_FDT 221539 2011-05-06 19:10:27Z ru Set to not build Flattened Device Tree support as part of the base system. This includes the device tree compiler (dtc) and libfdt support library. .It Va WITHOUT_FILE .\" from FreeBSD: head/tools/build/options/WITHOUT_FILE 278193 2015-02-04 10:24:40Z ngie Set to not build .Xr file 1 and related programs. .It Va WITHOUT_FINGER .\" from FreeBSD: head/tools/build/options/WITHOUT_FINGER 278192 2015-02-04 10:19:32Z ngie Set to not build or install .Xr finger 1 and .Xr fingerd 8 . .It Va WITHOUT_FLOPPY .\" from FreeBSD: head/tools/build/options/WITHOUT_FLOPPY 221540 2011-05-06 19:13:03Z ru Set to not build or install programs for operating floppy disk driver. .It Va WITH_FMAKE .\" from FreeBSD: head/tools/build/options/WITH_FMAKE 275138 2014-11-26 20:43:09Z gjb Causes the old FreeBSD .Xr make 1 program to be built and installed as fmake. .It Va WITHOUT_FMTREE .\" from FreeBSD: head/tools/build/options/WITHOUT_FMTREE 261299 2014-01-30 21:37:43Z brooks Set to not build and install .Pa /usr/sbin/fmtree . .It Va WITHOUT_FORMAT_EXTENSIONS .\" from FreeBSD: head/tools/build/options/WITHOUT_FORMAT_EXTENSIONS 250658 2013-05-15 13:04:10Z brooks Set to not enable .Fl fformat-extensions when compiling the kernel. Also disables all format checking. .It Va WITHOUT_FORTH .\" from FreeBSD: head/tools/build/options/WITHOUT_FORTH 156932 2006-03-21 07:50:50Z ru Set to build bootloaders without Forth support. .It Va WITHOUT_FP_LIBC .\" from FreeBSD: head/tools/build/options/WITHOUT_FP_LIBC 156932 2006-03-21 07:50:50Z ru Set to build .Nm libc without floating-point support. .It Va WITHOUT_FREEBSD_UPDATE .\" from FreeBSD: head/tools/build/options/WITHOUT_FREEBSD_UPDATE 183242 2008-09-21 22:02:26Z sam Set to not build .Xr freebsd-update 8 . .It Va WITHOUT_FTP .\" from FreeBSD: head/tools/build/options/WITHOUT_FTP 278192 2015-02-04 10:19:32Z ngie Set to not build or install .Xr ftp 1 and .Xr ftpd 8 . .It Va WITHOUT_GAMES .\" from FreeBSD: head/tools/build/options/WITHOUT_GAMES 156932 2006-03-21 07:50:50Z ru Set to not build games. .It Va WITHOUT_GCC .\" from FreeBSD: head/tools/build/options/WITHOUT_GCC 264660 2014-04-18 17:03:58Z imp Set to not build and install gcc and g++ as part of the normal build process. .Pp It is a default setting on amd64/amd64, arm/arm, arm/armv6, arm/armv6hf, i386/i386 and pc98/i386. .It Va WITH_GCC .\" from FreeBSD: head/tools/build/options/WITH_GCC 255326 2013-09-06 20:49:48Z zeising Set to build and install gcc and g++. .Pp It is a default setting on arm/armeb, mips/mipsel, mips/mips, mips/mips64el, mips/mips64, mips/mipsn32, powerpc/powerpc, powerpc/powerpc64 and sparc64/sparc64. .It Va WITHOUT_GCC_BOOTSTRAP .\" from FreeBSD: head/tools/build/options/WITHOUT_GCC_BOOTSTRAP 273177 2014-10-16 18:28:11Z skreuzer Set to not build gcc and g++ as part of the bootstrap process. You must enable either gcc or clang bootstrap to be able to build the system, unless an alternative compiler is provided via XCC. .Pp It is a default setting on amd64/amd64, arm/arm, arm/armv6, arm/armv6hf, i386/i386 and pc98/i386. .It Va WITH_GCC_BOOTSTRAP .\" from FreeBSD: head/tools/build/options/WITH_GCC_BOOTSTRAP 264660 2014-04-18 17:03:58Z imp Set to build gcc and g++ as part of the bootstrap process. .Pp It is a default setting on arm/armeb, mips/mipsel, mips/mips, mips/mips64el, mips/mips64, mips/mipsn32, powerpc/powerpc, powerpc/powerpc64 and sparc64/sparc64. .It Va WITHOUT_GCOV .\" from FreeBSD: head/tools/build/options/WITHOUT_GCOV 156932 2006-03-21 07:50:50Z ru Set to not build the .Xr gcov 1 tool. .It Va WITHOUT_GDB .\" from FreeBSD: head/tools/build/options/WITHOUT_GDB 156932 2006-03-21 07:50:50Z ru Set to not build .Xr gdb 1 . .It Va WITHOUT_GNU .\" from FreeBSD: head/tools/build/options/WITHOUT_GNU 174550 2007-12-12 16:43:17Z ru Set to not build contributed GNU software as a part of the base system. This option can be useful if the system built must not contain any code covered by the GNU Public License due to legal reasons. .Bf -symbolic The option has no effect yet. .Ef When set, it also enforces the following options: .Pp .Bl -item -compact .It .Va WITHOUT_GNU_SUPPORT .El .It Va WITHOUT_GNUCXX .\" from FreeBSD: head/tools/build/options/WITHOUT_GNUCXX 255321 2013-09-06 20:08:03Z theraven Do not build the GNU C++ stack (g++, libstdc++). This is the default on platforms where clang is the system compiler. .Pp It is a default setting on amd64/amd64, arm/arm, arm/armv6, arm/armv6hf, i386/i386 and pc98/i386. .It Va WITH_GNUCXX .\" from FreeBSD: head/tools/build/options/WITH_GNUCXX 255321 2013-09-06 20:08:03Z theraven Build the GNU C++ stack (g++, libstdc++). This is the default on platforms where gcc is the system compiler. .Pp It is a default setting on arm/armeb, mips/mipsel, mips/mips, mips/mips64el, mips/mips64, mips/mipsn32, powerpc/powerpc, powerpc/powerpc64 and sparc64/sparc64. .It Va WITHOUT_GNU_GREP_COMPAT .\" from FreeBSD: head/tools/build/options/WITHOUT_GNU_GREP_COMPAT 273421 2014-10-21 20:44:33Z emaste Set this option to omit the gnu extensions to grep from being included in BSD grep. .It Va WITHOUT_GNU_SUPPORT .\" from FreeBSD: head/tools/build/options/WITHOUT_GNU_SUPPORT 156932 2006-03-21 07:50:50Z ru Set to build some programs without optional GNU support. .It Va WITHOUT_GPIO .\" from FreeBSD: head/tools/build/options/WITHOUT_GPIO 228081 2011-11-28 17:54:34Z dim Set to not build .Xr gpioctl 8 as part of the base system. .It Va WITHOUT_GPL_DTC .\" from FreeBSD: head/tools/build/options/WITHOUT_GPL_DTC 264515 2014-04-15 20:41:55Z imp Set to build the BSD licensed version of the device tree compiler, instead of the GPL'd one from elinux.org. .It Va WITHOUT_GROFF .\" from FreeBSD: head/tools/build/options/WITHOUT_GROFF 218941 2011-02-22 08:13:49Z uqs Set to not build .Xr groff 1 and .Xr vgrind 1 . You should consider installing the textproc/groff port to not break .Xr man 1 . .It Va WITHOUT_GSSAPI .\" from FreeBSD: head/tools/build/options/WITHOUT_GSSAPI 174548 2007-12-12 16:39:32Z ru Set to not build libgssapi. .It Va WITHOUT_HAST .\" from FreeBSD: head/tools/build/options/WITHOUT_HAST 277725 2015-01-26 06:27:07Z ngie Set to not build .Xr hastd 8 and related utilities. .It Va WITH_HESIOD .\" from FreeBSD: head/tools/build/options/WITH_HESIOD 156932 2006-03-21 07:50:50Z ru Set to build Hesiod support. .It Va WITHOUT_HTML .\" from FreeBSD: head/tools/build/options/WITHOUT_HTML 156932 2006-03-21 07:50:50Z ru Set to not build HTML docs. .It Va WITHOUT_HYPERV .\" from FreeBSD: head/tools/build/options/WITHOUT_HYPERV 271493 2014-09-13 02:15:31Z delphij Set to not build or install HyperV utilities. .It Va WITHOUT_ICONV .\" from FreeBSD: head/tools/build/options/WITHOUT_ICONV 254919 2013-08-26 17:15:56Z antoine Set to not build iconv as part of libc. .It Va WITHOUT_INCLUDES .\" from FreeBSD: head/tools/build/options/WITHOUT_INCLUDES 275138 2014-11-26 20:43:09Z gjb Set to not install header files. This option used to be spelled .Va NO_INCS . .Bf -symbolic The option does not work for build targets. .Ef .It Va WITHOUT_INET .\" from FreeBSD: head/tools/build/options/WITHOUT_INET 221266 2011-04-30 17:58:28Z bz Set to not build programs and libraries related to IPv4 networking. When set, it also enforces the following options: .Pp .Bl -item -compact .It .Va WITHOUT_INET_SUPPORT .El .It Va WITHOUT_INET6 .\" from FreeBSD: head/tools/build/options/WITHOUT_INET6 156932 2006-03-21 07:50:50Z ru Set to not build programs and libraries related to IPv6 networking. When set, it also enforces the following options: .Pp .Bl -item -compact .It .Va WITHOUT_INET6_SUPPORT .El .It Va WITHOUT_INET6_SUPPORT .\" from FreeBSD: head/tools/build/options/WITHOUT_INET6_SUPPORT 156932 2006-03-21 07:50:50Z ru Set to build libraries, programs, and kernel modules without IPv6 support. .It Va WITHOUT_INETD .\" from FreeBSD: head/tools/build/options/WITHOUT_INETD 278192 2015-02-04 10:19:32Z ngie Set to not build .Xr inetd 8 . .It Va WITHOUT_INET_SUPPORT .\" from FreeBSD: head/tools/build/options/WITHOUT_INET_SUPPORT 221266 2011-04-30 17:58:28Z bz Set to build libraries, programs, and kernel modules without IPv4 support. .It Va WITHOUT_INSTALLLIB .\" from FreeBSD: head/tools/build/options/WITHOUT_INSTALLLIB 174497 2007-12-09 21:56:21Z dougb Set this if you do not want to install optional libraries. For example when creating a .Xr nanobsd 8 image. .It Va WITH_INSTALL_AS_USER .\" from FreeBSD: head/tools/build/options/WITH_INSTALL_AS_USER 238021 2012-07-02 20:24:01Z marcel Set to make install targets succeed for non-root users by installing files with owner and group attributes set to that of the user running the .Xr make 1 command. The user still has to set the .Va DESTDIR variable to point to a directory where the user has write permissions. .It Va WITHOUT_IPFILTER .\" from FreeBSD: head/tools/build/options/WITHOUT_IPFILTER 156932 2006-03-21 07:50:50Z ru Set to not build IP Filter package. .It Va WITHOUT_IPFW .\" from FreeBSD: head/tools/build/options/WITHOUT_IPFW 183242 2008-09-21 22:02:26Z sam Set to not build IPFW tools. .It Va WITHOUT_ISCSI .\" from FreeBSD: head/tools/build/options/WITHOUT_ISCSI 277675 2015-01-25 04:20:11Z ngie Set to not build .Xr iscid 8 and related utilities. .It Va WITHOUT_JAIL .\" from FreeBSD: head/tools/build/options/WITHOUT_JAIL 249966 2013-04-27 04:09:09Z eadler Set to not build tools for the support of jails; e.g., .Xr jail 8 . .It Va WITHOUT_KDUMP .\" from FreeBSD: head/tools/build/options/WITHOUT_KDUMP 240690 2012-09-19 11:38:37Z zeising Set to not build .Xr kdump 1 and .Xr truss 1 . .It Va WITHOUT_KERBEROS .\" from FreeBSD: head/tools/build/options/WITHOUT_KERBEROS 174549 2007-12-12 16:42:03Z ru Set this if you do not want to build Kerberos 5 (KTH Heimdal). When set, it also enforces the following options: .Pp .Bl -item -compact .It .Va WITHOUT_KERBEROS_SUPPORT .El .Pp When set, the following options are also in effect: .Pp .Bl -inset -compact .It Va WITHOUT_GSSAPI (unless .Va WITH_GSSAPI is set explicitly) .El .It Va WITHOUT_KERBEROS_SUPPORT .\" from FreeBSD: head/tools/build/options/WITHOUT_KERBEROS_SUPPORT 251794 2013-06-15 20:29:07Z eadler Set to build some programs without Kerberos support, like .Xr ssh 1 , .Xr telnet 1 , .Xr sshd 8 , and .Xr telnetd 8 . .It Va WITHOUT_KERNEL_SYMBOLS .\" from FreeBSD: head/tools/build/options/WITHOUT_KERNEL_SYMBOLS 222189 2011-05-22 18:23:17Z imp Set to not install kernel symbol files. .Bf -symbolic This option is recommended for those people who have small root partitions. .Ef .It Va WITHOUT_KVM .\" from FreeBSD: head/tools/build/options/WITHOUT_KVM 174550 2007-12-12 16:43:17Z ru Set to not build the .Nm libkvm library as a part of the base system. .Bf -symbolic The option has no effect yet. .Ef When set, it also enforces the following options: .Pp .Bl -item -compact .It .Va WITHOUT_KVM_SUPPORT .El .It Va WITHOUT_KVM_SUPPORT .\" from FreeBSD: head/tools/build/options/WITHOUT_KVM_SUPPORT 170644 2007-06-13 02:08:04Z sepotvin Set to build some programs without optional .Nm libkvm support. .It Va WITHOUT_LDNS .\" from FreeBSD: head/tools/build/options/WITHOUT_LDNS 255591 2013-09-15 13:11:13Z des Setting this variable will prevent the LDNS library from being built. When set, it also enforces the following options: .Pp .Bl -item -compact .It .Va WITHOUT_LDNS_UTILS .It .Va WITHOUT_UNBOUND .El .It Va WITHOUT_LDNS_UTILS .\" from FreeBSD: head/tools/build/options/WITHOUT_LDNS_UTILS 255850 2013-09-24 14:33:31Z des Setting this variable will prevent building the LDNS utilities .Xr drill 1 and .Xr host 1 . .It Va WITHOUT_LEGACY_CONSOLE .\" from FreeBSD: head/tools/build/options/WITHOUT_LEGACY_CONSOLE 249966 2013-04-27 04:09:09Z eadler Set to not build programs that support a legacy PC console; e.g., .Xr kbdcontrol 8 and .Xr vidcontrol 8 . .It Va WITHOUT_LIB32 .\" from FreeBSD: head/tools/build/options/WITHOUT_LIB32 274664 2014-11-18 17:06:48Z imp On 64-bit platforms, set to not build 32-bit library set and a .Nm ld-elf32.so.1 runtime linker. .It Va WITHOUT_LIBCPLUSPLUS .\" from FreeBSD: head/tools/build/options/WITHOUT_LIBCPLUSPLUS 246262 2013-02-02 22:42:46Z dim Set to avoid building libcxxrt and libc++. .It Va WITHOUT_LIBPTHREAD .\" from FreeBSD: head/tools/build/options/WITHOUT_LIBPTHREAD 188848 2009-02-20 11:09:55Z mtm Set to not build the .Nm libpthread providing library, .Nm libthr . When set, it also enforces the following options: .Pp .Bl -item -compact .It .Va WITHOUT_LIBTHR .El .It Va WITHOUT_LIBTHR .\" from FreeBSD: head/tools/build/options/WITHOUT_LIBTHR 156932 2006-03-21 07:50:50Z ru Set to not build the .Nm libthr (1:1 threading) library. .It Va WITH_LLDB .\" from FreeBSD: head/tools/build/options/WITH_LLDB 255722 2013-09-20 01:52:02Z emaste Set to build the LLDB debugger. .It Va WITHOUT_LOCALES .\" from FreeBSD: head/tools/build/options/WITHOUT_LOCALES 156932 2006-03-21 07:50:50Z ru Set to not build localization files; see .Xr locale 1 . .It Va WITHOUT_LOCATE .\" from FreeBSD: head/tools/build/options/WITHOUT_LOCATE 183242 2008-09-21 22:02:26Z sam Set to not build .Xr locate 1 and related programs. .It Va WITHOUT_LPR .\" from FreeBSD: head/tools/build/options/WITHOUT_LPR 156932 2006-03-21 07:50:50Z ru Set to not build .Xr lpr 1 and related programs. .It Va WITHOUT_LS_COLORS .\" from FreeBSD: head/tools/build/options/WITHOUT_LS_COLORS 235660 2012-05-19 20:05:27Z marcel Set to build .Xr ls 1 without support for colors to distinguish file types. .It Va WITHOUT_LZMA_SUPPORT .\" from FreeBSD: head/tools/build/options/WITHOUT_LZMA_SUPPORT 245171 2013-01-08 18:37:12Z obrien Set to build some programs without optional lzma compression support. .It Va WITHOUT_MAIL .\" from FreeBSD: head/tools/build/options/WITHOUT_MAIL 183242 2008-09-21 22:02:26Z sam Set to not build any mail support (MUA or MTA). When set, it also enforces the following options: .Pp .Bl -item -compact .It .Va WITHOUT_DMAGENT .It .Va WITHOUT_MAILWRAPPER .It .Va WITHOUT_SENDMAIL .El .It Va WITHOUT_MAILWRAPPER .\" from FreeBSD: head/tools/build/options/WITHOUT_MAILWRAPPER 156932 2006-03-21 07:50:50Z ru Set to not build the .Xr mailwrapper 8 MTA selector. .It Va WITHOUT_MAKE .\" from FreeBSD: head/tools/build/options/WITHOUT_MAKE 183242 2008-09-21 22:02:26Z sam Set to not install .Xr make 1 and related support files. .It Va WITHOUT_MAN .\" from FreeBSD: head/tools/build/options/WITHOUT_MAN 156932 2006-03-21 07:50:50Z ru Set to not build manual pages. When set, the following options are also in effect: .Pp .Bl -inset -compact .It Va WITHOUT_MAN_UTILS (unless .Va WITH_MAN_UTILS is set explicitly) .El .It Va WITHOUT_MANCOMPRESS .\" from FreeBSD: head/tools/build/options/WITHOUT_MANCOMPRESS 266752 2014-05-27 15:52:27Z gjb Set to not to install compressed man pages. Only the uncompressed versions will be installed. +.It Va WITHOUT_MANDOCDB +.\" from FreeBSD: head/tools/build/options/WITHOUT_MANDOCDB 283777 2015-05-30 17:41:37Z bapt +Use the +.Xr mandoc 1 +version of +.Xr makewhatis 8 +database and utilities. .It Va WITHOUT_MAN_UTILS .\" from FreeBSD: head/tools/build/options/WITHOUT_MAN_UTILS 208322 2010-05-20 00:07:21Z jkim Set to not build utilities for manual pages, .Xr apropos 1 , .Xr catman 1 , .Xr makewhatis 1 , .Xr man 1 , .Xr whatis 1 , .Xr manctl 8 , and related support files. .It Va WITH_NAND .\" from FreeBSD: head/tools/build/options/WITH_NAND 235537 2012-05-17 10:11:18Z gber Set to build the NAND Flash components. .It Va WITHOUT_NDIS .\" from FreeBSD: head/tools/build/options/WITHOUT_NDIS 183242 2008-09-21 22:02:26Z sam Set to not build programs and libraries related to NDIS emulation support. .It Va WITHOUT_NETCAT .\" from FreeBSD: head/tools/build/options/WITHOUT_NETCAT 156932 2006-03-21 07:50:50Z ru Set to not build .Xr nc 1 utility. .It Va WITHOUT_NETGRAPH .\" from FreeBSD: head/tools/build/options/WITHOUT_NETGRAPH 183242 2008-09-21 22:02:26Z sam Set to not build applications to support .Xr netgraph 4 . When set, it also enforces the following options: .Pp .Bl -item -compact .It .Va WITHOUT_ATM .It .Va WITHOUT_BLUETOOTH .It .Va WITHOUT_NETGRAPH_SUPPORT .El .It Va WITHOUT_NETGRAPH_SUPPORT .\" from FreeBSD: head/tools/build/options/WITHOUT_NETGRAPH_SUPPORT 183305 2008-09-23 16:11:15Z sam Set to build libraries, programs, and kernel modules without netgraph support. .It Va WITHOUT_NIS .\" from FreeBSD: head/tools/build/options/WITHOUT_NIS 156932 2006-03-21 07:50:50Z ru Set to not build .Xr NIS 8 support and related programs. If set, you might need to adopt your .Xr nsswitch.conf 5 and remove .Sq nis entries. .It Va WITHOUT_NLS .\" from FreeBSD: head/tools/build/options/WITHOUT_NLS 156932 2006-03-21 07:50:50Z ru Set to not build NLS catalogs. .It Va WITHOUT_NLS_CATALOGS .\" from FreeBSD: head/tools/build/options/WITHOUT_NLS_CATALOGS 156932 2006-03-21 07:50:50Z ru Set to not build NLS catalog support for .Xr csh 1 . .It Va WITHOUT_NS_CACHING .\" from FreeBSD: head/tools/build/options/WITHOUT_NS_CACHING 172803 2007-10-19 14:01:25Z ru Set to disable name caching in the .Pa nsswitch subsystem. The generic caching daemon, .Xr nscd 8 , will not be built either if this option is set. .It Va WITHOUT_NTP .\" from FreeBSD: head/tools/build/options/WITHOUT_NTP 183242 2008-09-21 22:02:26Z sam Set to not build .Xr ntpd 8 and related programs. .It Va WITH_OFED .\" from FreeBSD: head/tools/build/options/WITH_OFED 228081 2011-11-28 17:54:34Z dim Set to build the .Dq "OpenFabrics Enterprise Distribution" Infiniband software stack. .It Va WITH_OPENLDAP .\" from FreeBSD: head/tools/build/options/WITH_OPENLDAP 264902 2014-04-24 23:17:31Z imp Enable building openldap support for kerberos. .It Va WITHOUT_OPENSSH .\" from FreeBSD: head/tools/build/options/WITHOUT_OPENSSH 156932 2006-03-21 07:50:50Z ru Set to not build OpenSSH. .It Va WITH_OPENSSH_NONE_CIPHER .\" from FreeBSD: head/tools/build/options/WITH_OPENSSH_NONE_CIPHER 245527 2013-01-17 01:51:04Z bz Set to include the "None" cipher support in OpenSSH and its libraries. Additional adjustments may need to be done to system configuration files, such as .Xr sshd_config 5 , to enable this cipher. Please see .Pa /usr/src/crypto/openssh/README.hpn for full details. .It Va WITHOUT_OPENSSL .\" from FreeBSD: head/tools/build/options/WITHOUT_OPENSSL 156932 2006-03-21 07:50:50Z ru Set to not build OpenSSL. When set, it also enforces the following options: .Pp .Bl -item -compact .It .Va WITHOUT_KERBEROS .It .Va WITHOUT_KERBEROS_SUPPORT .It .Va WITHOUT_OPENSSH .El .Pp When set, the following options are also in effect: .Pp .Bl -inset -compact .It Va WITHOUT_GSSAPI (unless .Va WITH_GSSAPI is set explicitly) .El .It Va WITHOUT_PAM .\" from FreeBSD: head/tools/build/options/WITHOUT_PAM 174550 2007-12-12 16:43:17Z ru Set to not build PAM library and modules. .Bf -symbolic This option is deprecated and does nothing. .Ef When set, it also enforces the following options: .Pp .Bl -item -compact .It .Va WITHOUT_PAM_SUPPORT .El .It Va WITHOUT_PAM_SUPPORT .\" from FreeBSD: head/tools/build/options/WITHOUT_PAM_SUPPORT 156932 2006-03-21 07:50:50Z ru Set to build some programs without PAM support, particularly .Xr ftpd 8 and .Xr ppp 8 . .It Va WITHOUT_PC_SYSINSTALL .\" from FreeBSD: head/tools/build/options/WITHOUT_PC_SYSINSTALL 245606 2013-01-18 15:57:09Z eadler Set to not build .Xr pc-sysinstall 8 and related programs. .It Va WITHOUT_PF .\" from FreeBSD: head/tools/build/options/WITHOUT_PF 156932 2006-03-21 07:50:50Z ru Set to not build PF firewall package. When set, it also enforces the following options: .Pp .Bl -item -compact .It .Va WITHOUT_AUTHPF .El .It Va WITHOUT_PKGBOOTSTRAP .\" from FreeBSD: head/tools/build/options/WITHOUT_PKGBOOTSTRAP 258924 2013-12-04 15:58:42Z bdrewery Set to not build .Xr pkg 7 bootstrap tool. .It Va WITHOUT_PMC .\" from FreeBSD: head/tools/build/options/WITHOUT_PMC 183242 2008-09-21 22:02:26Z sam Set to not build .Xr pmccontrol 8 and related programs. .It Va WITHOUT_PORTSNAP .\" from FreeBSD: head/tools/build/options/WITHOUT_PORTSNAP 183242 2008-09-21 22:02:26Z sam Set to not build or install .Xr portsnap 8 and related files. .It Va WITHOUT_PPP .\" from FreeBSD: head/tools/build/options/WITHOUT_PPP 183242 2008-09-21 22:02:26Z sam Set to not build .Xr ppp 8 and related programs. .It Va WITHOUT_PROFILE .\" from FreeBSD: head/tools/build/options/WITHOUT_PROFILE 228196 2011-12-02 09:09:54Z fjoe Set to avoid compiling profiled libraries. .It Va WITHOUT_QUOTAS .\" from FreeBSD: head/tools/build/options/WITHOUT_QUOTAS 183242 2008-09-21 22:02:26Z sam Set to not build .Xr quota 8 and related programs. .It Va WITHOUT_RADIUS_SUPPORT .\" from FreeBSD: head/tools/build/options/WITHOUT_RADIUS_SUPPORT 278182 2015-02-04 06:53:45Z ngie Set to not build radius support into various applications, like .Xr pam_radius 8 and .Xr ppp 8 . .It Va WITHOUT_RBOOTD .\" from FreeBSD: head/tools/build/options/WITHOUT_RBOOTD 278192 2015-02-04 10:19:32Z ngie Set to not build or install .Xr rbootd 8 . .It Va WITHOUT_RCMDS .\" from FreeBSD: head/tools/build/options/WITHOUT_RCMDS 156932 2006-03-21 07:50:50Z ru Disable building of the .Bx r-commands. This includes .Xr rlogin 1 , .Xr rsh 1 , etc. .It Va WITHOUT_RCS .\" from FreeBSD: head/tools/build/options/WITHOUT_RCS 275138 2014-11-26 20:43:09Z gjb Set to not build .Xr rcs 1 , .Xr etcupdate 8 , and related utilities. .It Va WITHOUT_RESCUE .\" from FreeBSD: head/tools/build/options/WITHOUT_RESCUE 156932 2006-03-21 07:50:50Z ru Set to not build .Xr rescue 8 . .It Va WITHOUT_ROUTED .\" from FreeBSD: head/tools/build/options/WITHOUT_ROUTED 183242 2008-09-21 22:02:26Z sam Set to not build .Xr routed 8 utility. .It Va WITHOUT_SENDMAIL .\" from FreeBSD: head/tools/build/options/WITHOUT_SENDMAIL 156932 2006-03-21 07:50:50Z ru Set to not build .Xr sendmail 8 and related programs. .It Va WITHOUT_SETUID_LOGIN .\" from FreeBSD: head/tools/build/options/WITHOUT_SETUID_LOGIN 156932 2006-03-21 07:50:50Z ru Set this to disable the installation of .Xr login 1 as a set-user-ID root program. .It Va WITHOUT_SHAREDOCS .\" from FreeBSD: head/tools/build/options/WITHOUT_SHAREDOCS 156932 2006-03-21 07:50:50Z ru Set to not build the .Bx 4.4 legacy docs. .It Va WITH_SHARED_TOOLCHAIN .\" from FreeBSD: head/tools/build/options/WITH_SHARED_TOOLCHAIN 235342 2012-05-12 16:12:36Z gjb Set to build the toolchain binaries shared. The set includes .Xr cc 1 , .Xr make 1 and necessary utilities like assembler, linker and library archive manager. .It Va WITH_SORT_THREADS .\" from FreeBSD: head/tools/build/options/WITH_SORT_THREADS 264158 2014-04-05 18:00:45Z imp Set to enable threads in .Xr sort 1 . .It Va WITHOUT_SOURCELESS .\" from FreeBSD: head/tools/build/options/WITHOUT_SOURCELESS 230972 2012-02-04 00:54:43Z rmh Set to not build kernel modules that include sourceless code (either microcode or native code for host CPU). When set, it also enforces the following options: .Pp .Bl -item -compact .It .Va WITHOUT_SOURCELESS_HOST .It .Va WITHOUT_SOURCELESS_UCODE .El .It Va WITHOUT_SOURCELESS_HOST .\" from FreeBSD: head/tools/build/options/WITHOUT_SOURCELESS_HOST 230972 2012-02-04 00:54:43Z rmh Set to not build kernel modules that include sourceless native code for host CPU. .It Va WITHOUT_SOURCELESS_UCODE .\" from FreeBSD: head/tools/build/options/WITHOUT_SOURCELESS_UCODE 230972 2012-02-04 00:54:43Z rmh Set to not build kernel modules that include sourceless microcode. .It Va WITHOUT_SSP .\" from FreeBSD: head/tools/build/options/WITHOUT_SSP 180012 2008-06-25 21:33:28Z ru Set to not build world with propolice stack smashing protection. .It Va WITH_SVN .\" from FreeBSD: head/tools/build/options/WITH_SVN 252561 2013-07-03 12:36:47Z zeising Set to install .Xr svnlite 1 as .Xr svn 1 . .It Va WITHOUT_SVNLITE .\" from FreeBSD: head/tools/build/options/WITHOUT_SVNLITE 252561 2013-07-03 12:36:47Z zeising Set to not build .Xr svnlite 1 and related programs. .It Va WITHOUT_SYMVER .\" from FreeBSD: head/tools/build/options/WITHOUT_SYMVER 169649 2007-05-17 05:03:24Z deischen Set to disable symbol versioning when building shared libraries. .It Va WITHOUT_SYSCONS .\" from FreeBSD: head/tools/build/options/WITHOUT_SYSCONS 156932 2006-03-21 07:50:50Z ru Set to not build .Xr syscons 4 support files such as keyboard maps, fonts, and screen output maps. .It Va WITHOUT_SYSINSTALL .\" from FreeBSD: head/tools/build/options/WITHOUT_SYSINSTALL 183242 2008-09-21 22:02:26Z sam Set to not build .Xr sysinstall 8 and related programs. .It Va WITHOUT_TALK .\" from FreeBSD: head/tools/build/options/WITHOUT_TALK 277676 2015-01-25 04:37:44Z ngie Set to not build or install .Xr talk 1 and .Xr talkd 8 . .It Va WITHOUT_TCP_WRAPPERS .\" from FreeBSD: head/tools/build/options/WITHOUT_TCP_WRAPPERS 278192 2015-02-04 10:19:32Z ngie Set to not build or install .Xr tcpd 8 , and related utilities. .It Va WITHOUT_TCSH .\" from FreeBSD: head/tools/build/options/WITHOUT_TCSH 156932 2006-03-21 07:50:50Z ru Set to not build and install .Pa /bin/csh (which is .Xr tcsh 1 ) . .It Va WITHOUT_TELNET .\" from FreeBSD: head/tools/build/options/WITHOUT_TELNET 183242 2008-09-21 22:02:26Z sam Set to not build .Xr telnet 8 and related programs. .It Va WITHOUT_TESTS .\" from FreeBSD: head/tools/build/options/WITHOUT_TESTS 268778 2014-07-16 21:40:11Z jmmv Set to not build nor install the .Fx Test Suite in .Pa /usr/tests/ . See .Xr tests 7 for more details. This also disables the build of all test-related dependencies, including ATF. When set, it also enforces the following options: .Pp .Bl -item -compact .It .Va WITHOUT_TESTS_SUPPORT .El .It Va WITHOUT_TESTS_SUPPORT .\" from FreeBSD: head/tools/build/options/WITHOUT_TESTS_SUPPORT 274665 2014-11-18 17:06:50Z imp Set to disables the build of all test-related dependencies, including ATF. .It Va WITHOUT_TEXTPROC .\" from FreeBSD: head/tools/build/options/WITHOUT_TEXTPROC 183242 2008-09-21 22:02:26Z sam Set to not build programs used for text processing. When set, it also enforces the following options: .Pp .Bl -item -compact .It .Va WITHOUT_GROFF .El .It Va WITHOUT_TFTP .\" from FreeBSD: head/tools/build/options/WITHOUT_TFTP 278192 2015-02-04 10:19:32Z ngie Set to not build or install .Xr tftp 1 and .Xr tftpd 8 . .It Va WITHOUT_TIMED .\" from FreeBSD: head/tools/build/options/WITHOUT_TIMED 278192 2015-02-04 10:19:32Z ngie Set to not build or install .Xr timed 8 . .It Va WITHOUT_TOOLCHAIN .\" from FreeBSD: head/tools/build/options/WITHOUT_TOOLCHAIN 273172 2014-10-16 15:55:13Z brooks Set to not install header or programs used for program development, compilers, debuggers etc. .Bf -symbolic The option does not work for build targets. .Ef When set, it also enforces the following options: .Pp .Bl -item -compact .It .Va WITHOUT_BINUTILS .It .Va WITHOUT_CLANG .It .Va WITHOUT_CLANG_EXTRAS .It .Va WITHOUT_CLANG_FULL .It .Va WITHOUT_GCC .It .Va WITHOUT_GDB .It .Va WITHOUT_INCLUDES .El .It Va WITHOUT_UNBOUND .\" from FreeBSD: head/tools/build/options/WITHOUT_UNBOUND 255597 2013-09-15 14:51:23Z des Set to not build .Xr unbound 8 and related programs. .It Va WITHOUT_USB .\" from FreeBSD: head/tools/build/options/WITHOUT_USB 156932 2006-03-21 07:50:50Z ru Set to not build USB-related programs and libraries. .It Va WITHOUT_USB_GADGET_EXAMPLES .\" from FreeBSD: head/tools/build/options/WITHOUT_USB_GADGET_EXAMPLES 274665 2014-11-18 17:06:50Z imp Set to build USB gadget kernel modules. .It Va WITHOUT_UTMPX .\" from FreeBSD: head/tools/build/options/WITHOUT_UTMPX 231530 2012-02-11 20:28:42Z ed Set to not build user accounting tools such as .Xr last 1 , .Xr users 1 , .Xr who 1 , .Xr ac 8 , .Xr lastlogin 8 and .Xr utx 8 . .It Va WITHOUT_VI .\" from FreeBSD: head/tools/build/options/WITHOUT_VI 264903 2014-04-24 23:17:40Z imp Set to not build and install vi, view, ex and related programs. .It Va WITHOUT_VT .\" from FreeBSD: head/tools/build/options/WITHOUT_VT 268022 2014-06-30 00:20:12Z emaste Set to not build .Xr vt 4 support files (fonts and keymaps). .It Va WITHOUT_WARNS .\" from FreeBSD: head/tools/build/options/WITHOUT_WARNS 276559 2015-01-02 18:57:58Z bapt Set this to not add warning flags to the compiler invocations. Useful as a temporary workaround when code enters the tree which triggers warnings in environments that differ from the original developer. .It Va WITHOUT_WIRELESS .\" from FreeBSD: head/tools/build/options/WITHOUT_WIRELESS 183242 2008-09-21 22:02:26Z sam Set to not build programs used for 802.11 wireless networks; especially .Xr wpa_supplicant 8 and .Xr hostapd 8 . When set, it also enforces the following options: .Pp .Bl -item -compact .It .Va WITHOUT_WIRELESS_SUPPORT .El .It Va WITHOUT_WIRELESS_SUPPORT .\" from FreeBSD: head/tools/build/options/WITHOUT_WIRELESS_SUPPORT 183305 2008-09-23 16:11:15Z sam Set to build libraries, programs, and kernel modules without 802.11 wireless support. .It Va WITHOUT_WPA_SUPPLICANT_EAPOL .\" from FreeBSD: head/tools/build/options/WITHOUT_WPA_SUPPLICANT_EAPOL 156932 2006-03-21 07:50:50Z ru Build .Xr wpa_supplicant 8 without support for the IEEE 802.1X protocol and without support for EAP-PEAP, EAP-TLS, EAP-LEAP, and EAP-TTLS protocols (usable only via 802.1X). .It Va WITHOUT_ZFS .\" from FreeBSD: head/tools/build/options/WITHOUT_ZFS 168409 2007-04-06 02:13:30Z pjd Set to not build ZFS file system. .It Va WITHOUT_ZONEINFO .\" from FreeBSD: head/tools/build/options/WITHOUT_ZONEINFO 235342 2012-05-12 16:12:36Z gjb Set to not build the timezone database. .El .Sh FILES .Bl -tag -compact -width Pa .It Pa /etc/src.conf .It Pa /usr/share/mk/bsd.own.mk .El .Sh SEE ALSO .Xr make 1 , .Xr make.conf 5 , .Xr build 7 , .Xr ports 7 .Sh HISTORY The .Nm file appeared in .Fx 7.0 . .Sh AUTHORS This manual page was autogenerated. Index: projects/clang-trunk/share/mk/src.opts.mk =================================================================== --- projects/clang-trunk/share/mk/src.opts.mk (revision 283870) +++ projects/clang-trunk/share/mk/src.opts.mk (revision 283871) @@ -1,398 +1,399 @@ # $FreeBSD$ # # Option file for FreeBSD /usr/src builds. # # Users define WITH_FOO and WITHOUT_FOO on the command line or in /etc/src.conf # and /etc/make.conf files. These translate in the build system to MK_FOO={yes,no} # with sensible (usually) defaults. # # Makefiles must include bsd.opts.mk after defining specific MK_FOO options that # are applicable for that Makefile (typically there are none, but sometimes there # are exceptions). Recursive makes usually add MK_FOO=no for options that they wish # to omit from that make. # # Makefiles must include bsd.mkopt.mk before they test the value of any MK_FOO # variable. # # Makefiles may also assume that this file is included by src.opts.mk should it # need variables defined there prior to the end of the Makefile where # bsd.{subdir,lib.bin}.mk is traditionally included. # # The old-style YES_FOO and NO_FOO are being phased out. No new instances of them # should be added. Old instances should be removed since they were just to # bridge the gap between FreeBSD 4 and FreeBSD 5. # # Makefiles should never test WITH_FOO or WITHOUT_FOO directly (although an # exception is made for _WITHOUT_SRCONF which turns off this mechanism # completely inside bsd.*.mk files). # .if !target(____) ____: .include # # Define MK_* variables (which are either "yes" or "no") for users # to set via WITH_*/WITHOUT_* in /etc/src.conf and override in the # make(1) environment. # These should be tested with `== "no"' or `!= "no"' in makefiles. # The NO_* variables should only be set by makefiles for variables # that haven't been converted over. # # These options are used by src the builds __DEFAULT_YES_OPTIONS = \ ACCT \ ACPI \ AMD \ APM \ AT \ ATM \ AUDIT \ AUTHPF \ AUTOFS \ BHYVE \ BINUTILS \ BINUTILS_BOOTSTRAP \ BLUETOOTH \ BOOT \ BOOTPARAMD \ BOOTPD \ BSD_CPIO \ BSDINSTALL \ BSNMP \ BZIP2 \ CALENDAR \ CAPSICUM \ CASPER \ CCD \ CDDL \ CPP \ CROSS_COMPILER \ CRYPT \ CTM \ CUSE \ CXX \ DICT \ DMAGENT \ DYNAMICROOT \ ED_CRYPTO \ EE \ ELFTOOLCHAIN_TOOLS \ EXAMPLES \ FDT \ FILE \ FINGER \ FLOPPY \ FMTREE \ FORTH \ FP_LIBC \ FREEBSD_UPDATE \ FTP \ GAMES \ GCOV \ GDB \ GNU \ GNU_GREP_COMPAT \ GPIO \ GPL_DTC \ GROFF \ HAST \ HTML \ HYPERV \ ICONV \ INET \ INET6 \ INETD \ IPFILTER \ IPFW \ ISCSI \ JAIL \ KDUMP \ KVM \ LDNS \ LDNS_UTILS \ LEGACY_CONSOLE \ LIB32 \ LIBPTHREAD \ LIBTHR \ LOCALES \ LOCATE \ LPR \ LS_COLORS \ LZMA_SUPPORT \ MAIL \ MAILWRAPPER \ MAKE \ + MANDOCDB \ NDIS \ NETCAT \ NETGRAPH \ NLS_CATALOGS \ NS_CACHING \ NTP \ OPENSSL \ PAM \ PC_SYSINSTALL \ PF \ PKGBOOTSTRAP \ PMC \ PORTSNAP \ PPP \ QUOTAS \ RADIUS_SUPPORT \ RCMDS \ RBOOTD \ RCS \ RESCUE \ ROUTED \ SENDMAIL \ SETUID_LOGIN \ SHAREDOCS \ SOURCELESS \ SOURCELESS_HOST \ SOURCELESS_UCODE \ SVNLITE \ SYSCONS \ SYSINSTALL \ TALK \ TCP_WRAPPERS \ TCSH \ TELNET \ TESTS \ TEXTPROC \ TFTP \ TIMED \ UNBOUND \ USB \ UTMPX \ VI \ VT \ WIRELESS \ WPA_SUPPLICANT_EAPOL \ ZFS \ ZONEINFO __DEFAULT_NO_OPTIONS = \ BSD_GREP \ CLANG_EXTRAS \ EISA \ FMAKE \ HESIOD \ LLDB \ NAND \ OFED \ OPENLDAP \ OPENSSH_NONE_CIPHER \ SHARED_TOOLCHAIN \ SORT_THREADS \ SVN # # Default behaviour of some options depends on the architecture. Unfortunately # this means that we have to test TARGET_ARCH (the buildworld case) as well # as MACHINE_ARCH (the non-buildworld case). Normally TARGET_ARCH is not # used at all in bsd.*.mk, but we have to make an exception here if we want # to allow defaults for some things like clang to vary by target architecture. # Additional, per-target behavior should be rarely added only after much # gnashing of teeth and grinding of gears. # .if defined(TARGET_ARCH) __T=${TARGET_ARCH} .else __T=${MACHINE_ARCH} .endif .if defined(TARGET) __TT=${TARGET} .else __TT=${MACHINE} .endif .include .if !${COMPILER_FEATURES:Mc++11} # If the compiler is not C++11 capable, disable clang and use gcc instead. __DEFAULT_YES_OPTIONS+=GCC GCC_BOOTSTRAP GNUCXX __DEFAULT_NO_OPTIONS+=CLANG CLANG_BOOTSTRAP CLANG_FULL CLANG_IS_CC .elif ${__T} == "aarch64" || ${__T} == "amd64" || ${__T} == "i386" # On x86 and arm64, clang is enabled, and will be installed as the default cc. __DEFAULT_YES_OPTIONS+=CLANG CLANG_BOOTSTRAP CLANG_FULL CLANG_IS_CC __DEFAULT_NO_OPTIONS+=GCC GCC_BOOTSTRAP GNUCXX .elif ${__TT} == "arm" && ${__T:Marm*eb*} == "" # On little-endian arm, clang is enabled, and it is installed as the default # cc, but since gcc is unable to build the full clang, disable it by default. __DEFAULT_YES_OPTIONS+=CLANG CLANG_BOOTSTRAP CLANG_IS_CC __DEFAULT_NO_OPTIONS+=CLANG_FULL GCC GCC_BOOTSTRAP GNUCXX .elif ${__T:Mpowerpc*} # On powerpc, clang is enabled, but gcc is installed as the default cc. __DEFAULT_YES_OPTIONS+=CLANG CLANG_FULL GCC GCC_BOOTSTRAP GNUCXX __DEFAULT_NO_OPTIONS+=CLANG_BOOTSTRAP CLANG_IS_CC .else # Everything else disables clang, and uses gcc instead. __DEFAULT_YES_OPTIONS+=GCC GCC_BOOTSTRAP GNUCXX __DEFAULT_NO_OPTIONS+=CLANG CLANG_BOOTSTRAP CLANG_FULL CLANG_IS_CC .endif .if ${__T} == "aarch64" BROKEN_OPTIONS+=BINUTILS BINUTILS_BOOTSTRAP GCC GCC_BOOTSTRAP GDB .endif # LLVM lacks support for FreeBSD 64-bit atomic operations for ARMv4/ARMv5 .if ${__T} == "arm" || ${__T} == "armeb" BROKEN_OPTIONS+=LLDB .endif .include # # MK_* options that default to "yes" if the compiler is a C++11 compiler. # .for var in \ LIBCPLUSPLUS .if !defined(MK_${var}) .if ${COMPILER_FEATURES:Mc++11} .if defined(WITHOUT_${var}) MK_${var}:= no .else MK_${var}:= yes .endif .else .if defined(WITH_${var}) MK_${var}:= yes .else MK_${var}:= no .endif .endif .endif .endfor # # Force some options off if their dependencies are off. # Order is somewhat important. # .if ${MK_LIBPTHREAD} == "no" MK_LIBTHR:= no .endif .if ${MK_LDNS} == "no" MK_LDNS_UTILS:= no MK_UNBOUND:= no .endif .if ${MK_SOURCELESS} == "no" MK_SOURCELESS_HOST:= no MK_SOURCELESS_UCODE:= no .endif .if ${MK_CDDL} == "no" MK_ZFS:= no MK_CTF:= no .endif .if ${MK_CRYPT} == "no" MK_OPENSSL:= no MK_OPENSSH:= no MK_KERBEROS:= no .endif .if ${MK_CXX} == "no" MK_CLANG:= no MK_GROFF:= no MK_GNUCXX:= no .endif .if ${MK_MAIL} == "no" MK_MAILWRAPPER:= no MK_SENDMAIL:= no MK_DMAGENT:= no .endif .if ${MK_NETGRAPH} == "no" MK_ATM:= no MK_BLUETOOTH:= no .endif .if ${MK_OPENSSL} == "no" MK_OPENSSH:= no MK_KERBEROS:= no .endif .if ${MK_PF} == "no" MK_AUTHPF:= no .endif .if ${MK_TEXTPROC} == "no" MK_GROFF:= no .endif .if ${MK_CROSS_COMPILER} == "no" MK_BINUTILS_BOOTSTRAP:= no MK_CLANG_BOOTSTRAP:= no MK_GCC_BOOTSTRAP:= no .endif .if ${MK_TOOLCHAIN} == "no" MK_BINUTILS:= no MK_CLANG:= no MK_GCC:= no MK_GDB:= no MK_INCLUDES:= no .endif .if ${MK_CLANG} == "no" MK_CLANG_EXTRAS:= no MK_CLANG_FULL:= no .endif # # Set defaults for the MK_*_SUPPORT variables. # # # MK_*_SUPPORT options which default to "yes" unless their corresponding # MK_* variable is set to "no". # .for var in \ BZIP2 \ GNU \ INET \ INET6 \ KERBEROS \ KVM \ NETGRAPH \ PAM \ TESTS \ WIRELESS .if defined(WITHOUT_${var}_SUPPORT) || ${MK_${var}} == "no" MK_${var}_SUPPORT:= no .else MK_${var}_SUPPORT:= yes .endif .endfor # # MK_* options whose default value depends on another option. # .for vv in \ GSSAPI/KERBEROS \ MAN_UTILS/MAN .if defined(WITH_${vv:H}) MK_${vv:H}:= yes .elif defined(WITHOUT_${vv:H}) MK_${vv:H}:= no .else MK_${vv:H}:= ${MK_${vv:T}} .endif .endfor .if !${COMPILER_FEATURES:Mc++11} MK_LLDB:= no .endif # gcc 4.8 and newer supports libc++, so suppress gnuc++ in that case. # while in theory we could build it with that, we don't want to do # that since it creates too much confusion for too little gain. .if ${COMPILER_TYPE} == "gcc" && ${COMPILER_VERSION} >= 40800 MK_GNUCXX:=no MK_GCC:=no .endif .endif # !target(____) Index: projects/clang-trunk/share =================================================================== --- projects/clang-trunk/share (revision 283870) +++ projects/clang-trunk/share (revision 283871) Property changes on: projects/clang-trunk/share ___________________________________________________________________ Modified: svn:mergeinfo ## -0,0 +0,1 ## Merged /head/share:r283771-283870 Index: projects/clang-trunk/sys/amd64/amd64/trap.c =================================================================== --- projects/clang-trunk/sys/amd64/amd64/trap.c (revision 283870) +++ projects/clang-trunk/sys/amd64/amd64/trap.c (revision 283871) @@ -1,993 +1,987 @@ /*- * Copyright (C) 1994, David Greenman * Copyright (c) 1990, 1993 * The Regents of the University of California. All rights reserved. * * This code is derived from software contributed to Berkeley by * the University of Utah, and William Jolitz. * * Redistribution and use in source and binary forms, with or without * modification, are permitted provided that the following conditions * are met: * 1. Redistributions of source code must retain the above copyright * notice, this list of conditions and the following disclaimer. * 2. Redistributions in binary form must reproduce the above copyright * notice, this list of conditions and the following disclaimer in the * documentation and/or other materials provided with the distribution. * 3. All advertising materials mentioning features or use of this software * must display the following acknowledgement: * This product includes software developed by the University of * California, Berkeley and its contributors. * 4. Neither the name of the University nor the names of its contributors * may be used to endorse or promote products derived from this software * without specific prior written permission. * * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF * SUCH DAMAGE. * * from: @(#)trap.c 7.4 (Berkeley) 5/13/91 */ #include __FBSDID("$FreeBSD$"); /* * AMD64 Trap and System call handling */ #include "opt_clock.h" #include "opt_cpu.h" #include "opt_hwpmc_hooks.h" #include "opt_isa.h" #include "opt_kdb.h" #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #ifdef HWPMC_HOOKS #include PMC_SOFT_DEFINE( , , page_fault, all); PMC_SOFT_DEFINE( , , page_fault, read); PMC_SOFT_DEFINE( , , page_fault, write); #endif #include #include #include #include #include #include #include #include #include #include #include #include #ifdef SMP #include #endif #include #ifdef KDTRACE_HOOKS #include #endif extern void __noinline trap(struct trapframe *frame); extern void trap_check(struct trapframe *frame); extern void syscall(struct trapframe *frame); void dblfault_handler(struct trapframe *frame); static int trap_pfault(struct trapframe *, int); static void trap_fatal(struct trapframe *, vm_offset_t); #define MAX_TRAP_MSG 32 static char *trap_msg[] = { "", /* 0 unused */ "privileged instruction fault", /* 1 T_PRIVINFLT */ "", /* 2 unused */ "breakpoint instruction fault", /* 3 T_BPTFLT */ "", /* 4 unused */ "", /* 5 unused */ "arithmetic trap", /* 6 T_ARITHTRAP */ "", /* 7 unused */ "", /* 8 unused */ "general protection fault", /* 9 T_PROTFLT */ "trace trap", /* 10 T_TRCTRAP */ "", /* 11 unused */ "page fault", /* 12 T_PAGEFLT */ "", /* 13 unused */ "alignment fault", /* 14 T_ALIGNFLT */ "", /* 15 unused */ "", /* 16 unused */ "", /* 17 unused */ "integer divide fault", /* 18 T_DIVIDE */ "non-maskable interrupt trap", /* 19 T_NMI */ "overflow trap", /* 20 T_OFLOW */ "FPU bounds check fault", /* 21 T_BOUND */ "FPU device not available", /* 22 T_DNA */ "double fault", /* 23 T_DOUBLEFLT */ "FPU operand fetch fault", /* 24 T_FPOPFLT */ "invalid TSS fault", /* 25 T_TSSFLT */ "segment not present fault", /* 26 T_SEGNPFLT */ "stack fault", /* 27 T_STKFLT */ "machine check trap", /* 28 T_MCHK */ "SIMD floating-point exception", /* 29 T_XMMFLT */ "reserved (unknown) fault", /* 30 T_RESERVED */ "", /* 31 unused (reserved) */ "DTrace pid return trap", /* 32 T_DTRACE_RET */ }; #ifdef KDB static int kdb_on_nmi = 1; SYSCTL_INT(_machdep, OID_AUTO, kdb_on_nmi, CTLFLAG_RWTUN, &kdb_on_nmi, 0, "Go to KDB on NMI"); #endif static int panic_on_nmi = 1; SYSCTL_INT(_machdep, OID_AUTO, panic_on_nmi, CTLFLAG_RWTUN, &panic_on_nmi, 0, "Panic on NMI"); static int prot_fault_translation; SYSCTL_INT(_machdep, OID_AUTO, prot_fault_translation, CTLFLAG_RWTUN, &prot_fault_translation, 0, "Select signal to deliver on protection fault"); static int uprintf_signal; SYSCTL_INT(_machdep, OID_AUTO, uprintf_signal, CTLFLAG_RWTUN, &uprintf_signal, 0, "Print debugging information on trap signal to ctty"); /* * Exception, fault, and trap interface to the FreeBSD kernel. * This common code is called from assembly language IDT gate entry * routines that prepare a suitable stack frame, and restore this * frame after the exception has been processed. */ void trap(struct trapframe *frame) { #ifdef KDTRACE_HOOKS struct reg regs; #endif struct thread *td = curthread; struct proc *p = td->td_proc; int i = 0, ucode = 0, code; u_int type; register_t addr = 0; ksiginfo_t ksi; PCPU_INC(cnt.v_trap); type = frame->tf_trapno; #ifdef SMP /* Handler for NMI IPIs used for stopping CPUs. */ if (type == T_NMI) { if (ipi_nmi_handler() == 0) goto out; } #endif /* SMP */ #ifdef KDB if (kdb_active) { kdb_reenter(); goto out; } #endif if (type == T_RESERVED) { trap_fatal(frame, 0); goto out; } #ifdef HWPMC_HOOKS /* * CPU PMCs interrupt using an NMI. If the PMC module is * active, pass the 'rip' value to the PMC module's interrupt * handler. A return value of '1' from the handler means that * the NMI was handled by it and we can return immediately. */ if (type == T_NMI && pmc_intr && (*pmc_intr)(PCPU_GET(cpuid), frame)) goto out; #endif if (type == T_MCHK) { mca_intr(); goto out; } if ((frame->tf_rflags & PSL_I) == 0) { /* * Buggy application or kernel code has disabled * interrupts and then trapped. Enabling interrupts * now is wrong, but it is better than running with * interrupts disabled until they are accidentally * enabled later. */ if (ISPL(frame->tf_cs) == SEL_UPL) uprintf( "pid %ld (%s): trap %d with interrupts disabled\n", (long)curproc->p_pid, curthread->td_name, type); else if (type != T_NMI && type != T_BPTFLT && type != T_TRCTRAP) { /* * XXX not quite right, since this may be for a * multiple fault in user mode. */ printf("kernel trap %d with interrupts disabled\n", type); /* * We shouldn't enable interrupts while holding a * spin lock. */ if (td->td_md.md_spinlock_count == 0) enable_intr(); } } code = frame->tf_err; if (ISPL(frame->tf_cs) == SEL_UPL) { /* user trap */ td->td_pticks = 0; td->td_frame = frame; addr = frame->tf_rip; if (td->td_ucred != p->p_ucred) cred_update_thread(td); switch (type) { case T_PRIVINFLT: /* privileged instruction fault */ i = SIGILL; ucode = ILL_PRVOPC; break; case T_BPTFLT: /* bpt instruction fault */ case T_TRCTRAP: /* trace trap */ enable_intr(); #ifdef KDTRACE_HOOKS if (type == T_BPTFLT) { fill_frame_regs(frame, ®s); if (dtrace_pid_probe_ptr != NULL && dtrace_pid_probe_ptr(®s) == 0) goto out; } #endif frame->tf_rflags &= ~PSL_T; i = SIGTRAP; ucode = (type == T_TRCTRAP ? TRAP_TRACE : TRAP_BRKPT); break; case T_ARITHTRAP: /* arithmetic trap */ ucode = fputrap_x87(); if (ucode == -1) goto userout; i = SIGFPE; break; case T_PROTFLT: /* general protection fault */ i = SIGBUS; ucode = BUS_OBJERR; break; case T_STKFLT: /* stack fault */ case T_SEGNPFLT: /* segment not present fault */ i = SIGBUS; ucode = BUS_ADRERR; break; case T_TSSFLT: /* invalid TSS fault */ i = SIGBUS; ucode = BUS_OBJERR; break; case T_ALIGNFLT: i = SIGBUS; ucode = BUS_ADRALN; break; case T_DOUBLEFLT: /* double fault */ default: i = SIGBUS; ucode = BUS_OBJERR; break; case T_PAGEFLT: /* page fault */ addr = frame->tf_addr; i = trap_pfault(frame, TRUE); if (i == -1) goto userout; if (i == 0) goto user; if (i == SIGSEGV) ucode = SEGV_MAPERR; else { if (prot_fault_translation == 0) { /* * Autodetect. * This check also covers the images * without the ABI-tag ELF note. */ if (SV_CURPROC_ABI() == SV_ABI_FREEBSD && p->p_osrel >= P_OSREL_SIGSEGV) { i = SIGSEGV; ucode = SEGV_ACCERR; } else { i = SIGBUS; ucode = BUS_PAGE_FAULT; } } else if (prot_fault_translation == 1) { /* * Always compat mode. */ i = SIGBUS; ucode = BUS_PAGE_FAULT; } else { /* * Always SIGSEGV mode. */ i = SIGSEGV; ucode = SEGV_ACCERR; } } break; case T_DIVIDE: /* integer divide fault */ ucode = FPE_INTDIV; i = SIGFPE; break; #ifdef DEV_ISA case T_NMI: /* machine/parity/power fail/"kitchen sink" faults */ if (isa_nmi(code) == 0) { #ifdef KDB /* * NMI can be hooked up to a pushbutton * for debugging. */ if (kdb_on_nmi) { printf ("NMI ... going to debugger\n"); kdb_trap(type, 0, frame); } #endif /* KDB */ goto userout; } else if (panic_on_nmi) panic("NMI indicates hardware failure"); break; #endif /* DEV_ISA */ case T_OFLOW: /* integer overflow fault */ ucode = FPE_INTOVF; i = SIGFPE; break; case T_BOUND: /* bounds check fault */ ucode = FPE_FLTSUB; i = SIGFPE; break; case T_DNA: /* transparent fault (due to context switch "late") */ KASSERT(PCB_USER_FPU(td->td_pcb), ("kernel FPU ctx has leaked")); fpudna(); goto userout; case T_FPOPFLT: /* FPU operand fetch fault */ ucode = ILL_COPROC; i = SIGILL; break; case T_XMMFLT: /* SIMD floating-point exception */ ucode = fputrap_sse(); if (ucode == -1) goto userout; i = SIGFPE; break; #ifdef KDTRACE_HOOKS case T_DTRACE_RET: enable_intr(); fill_frame_regs(frame, ®s); if (dtrace_return_probe_ptr != NULL && dtrace_return_probe_ptr(®s) == 0) goto out; break; #endif } } else { /* kernel trap */ KASSERT(cold || td->td_ucred != NULL, ("kernel trap doesn't have ucred")); switch (type) { case T_PAGEFLT: /* page fault */ (void) trap_pfault(frame, FALSE); goto out; case T_DNA: KASSERT(!PCB_USER_FPU(td->td_pcb), ("Unregistered use of FPU in kernel")); fpudna(); goto out; case T_ARITHTRAP: /* arithmetic trap */ case T_XMMFLT: /* SIMD floating-point exception */ case T_FPOPFLT: /* FPU operand fetch fault */ /* * For now, supporting kernel handler * registration for FPU traps is overkill. */ trap_fatal(frame, 0); goto out; case T_STKFLT: /* stack fault */ case T_PROTFLT: /* general protection fault */ case T_SEGNPFLT: /* segment not present fault */ if (td->td_intr_nesting_level != 0) break; /* * Invalid segment selectors and out of bounds * %rip's and %rsp's can be set up in user mode. * This causes a fault in kernel mode when the * kernel tries to return to user mode. We want * to get this fault so that we can fix the * problem here and not have to check all the * selectors and pointers when the user changes * them. */ if (frame->tf_rip == (long)doreti_iret) { frame->tf_rip = (long)doreti_iret_fault; goto out; } if (frame->tf_rip == (long)ld_ds) { frame->tf_rip = (long)ds_load_fault; goto out; } if (frame->tf_rip == (long)ld_es) { frame->tf_rip = (long)es_load_fault; goto out; } if (frame->tf_rip == (long)ld_fs) { frame->tf_rip = (long)fs_load_fault; goto out; } if (frame->tf_rip == (long)ld_gs) { frame->tf_rip = (long)gs_load_fault; goto out; } if (frame->tf_rip == (long)ld_gsbase) { frame->tf_rip = (long)gsbase_load_fault; goto out; } if (frame->tf_rip == (long)ld_fsbase) { frame->tf_rip = (long)fsbase_load_fault; goto out; } if (curpcb->pcb_onfault != NULL) { frame->tf_rip = (long)curpcb->pcb_onfault; goto out; } break; case T_TSSFLT: /* * PSL_NT can be set in user mode and isn't cleared * automatically when the kernel is entered. This * causes a TSS fault when the kernel attempts to * `iret' because the TSS link is uninitialized. We * want to get this fault so that we can fix the * problem here and not every time the kernel is * entered. */ if (frame->tf_rflags & PSL_NT) { frame->tf_rflags &= ~PSL_NT; goto out; } break; case T_TRCTRAP: /* trace trap */ /* * Ignore debug register trace traps due to * accesses in the user's address space, which * can happen under several conditions such as * if a user sets a watchpoint on a buffer and * then passes that buffer to a system call. * We still want to get TRCTRAPS for addresses * in kernel space because that is useful when * debugging the kernel. */ if (user_dbreg_trap()) { /* * Reset breakpoint bits because the * processor doesn't */ /* XXX check upper bits here */ load_dr6(rdr6() & 0xfffffff0); goto out; } /* * FALLTHROUGH (TRCTRAP kernel mode, kernel address) */ case T_BPTFLT: /* * If KDB is enabled, let it handle the debugger trap. * Otherwise, debugger traps "can't happen". */ #ifdef KDB if (kdb_trap(type, 0, frame)) goto out; #endif break; #ifdef DEV_ISA case T_NMI: /* machine/parity/power fail/"kitchen sink" faults */ if (isa_nmi(code) == 0) { #ifdef KDB /* * NMI can be hooked up to a pushbutton * for debugging. */ if (kdb_on_nmi) { printf ("NMI ... going to debugger\n"); kdb_trap(type, 0, frame); } #endif /* KDB */ goto out; } else if (panic_on_nmi == 0) goto out; /* FALLTHROUGH */ #endif /* DEV_ISA */ } trap_fatal(frame, 0); goto out; } /* Translate fault for emulators (e.g. Linux) */ if (*p->p_sysent->sv_transtrap) i = (*p->p_sysent->sv_transtrap)(i, type); ksiginfo_init_trap(&ksi); ksi.ksi_signo = i; ksi.ksi_code = ucode; ksi.ksi_trapno = type; ksi.ksi_addr = (void *)addr; if (uprintf_signal) { uprintf("pid %d comm %s: signal %d err %lx code %d type %d " "addr 0x%lx rsp 0x%lx rip 0x%lx " "<%02x %02x %02x %02x %02x %02x %02x %02x>\n", p->p_pid, p->p_comm, i, frame->tf_err, ucode, type, addr, frame->tf_rsp, frame->tf_rip, fubyte((void *)(frame->tf_rip + 0)), fubyte((void *)(frame->tf_rip + 1)), fubyte((void *)(frame->tf_rip + 2)), fubyte((void *)(frame->tf_rip + 3)), fubyte((void *)(frame->tf_rip + 4)), fubyte((void *)(frame->tf_rip + 5)), fubyte((void *)(frame->tf_rip + 6)), fubyte((void *)(frame->tf_rip + 7))); } KASSERT((read_rflags() & PSL_I) != 0, ("interrupts disabled")); trapsignal(td, &ksi); user: userret(td, frame); KASSERT(PCB_USER_FPU(td->td_pcb), ("Return from trap with kernel FPU ctx leaked")); userout: out: return; } /* * Ensure that we ignore any DTrace-induced faults. This function cannot * be instrumented, so it cannot generate such faults itself. */ void trap_check(struct trapframe *frame) { #ifdef KDTRACE_HOOKS if (dtrace_trap_func != NULL && (*dtrace_trap_func)(frame, frame->tf_trapno) != 0) return; #endif trap(frame); } static int trap_pfault(frame, usermode) struct trapframe *frame; int usermode; { vm_offset_t va; struct vmspace *vm; vm_map_t map; int rv = 0; vm_prot_t ftype; struct thread *td = curthread; struct proc *p = td->td_proc; vm_offset_t eva = frame->tf_addr; if (__predict_false((td->td_pflags & TDP_NOFAULTING) != 0)) { /* * Due to both processor errata and lazy TLB invalidation when * access restrictions are removed from virtual pages, memory * accesses that are allowed by the physical mapping layer may * nonetheless cause one spurious page fault per virtual page. * When the thread is executing a "no faulting" section that * is bracketed by vm_fault_{disable,enable}_pagefaults(), * every page fault is treated as a spurious page fault, * unless it accesses the same virtual address as the most * recent page fault within the same "no faulting" section. */ if (td->td_md.md_spurflt_addr != eva || (td->td_pflags & TDP_RESETSPUR) != 0) { /* * Do nothing to the TLB. A stale TLB entry is * flushed automatically by a page fault. */ td->td_md.md_spurflt_addr = eva; td->td_pflags &= ~TDP_RESETSPUR; return (0); } } else { /* * If we get a page fault while in a critical section, then * it is most likely a fatal kernel page fault. The kernel * is already going to panic trying to get a sleep lock to * do the VM lookup, so just consider it a fatal trap so the * kernel can print out a useful trap message and even get * to the debugger. * * If we get a page fault while holding a non-sleepable * lock, then it is most likely a fatal kernel page fault. * If WITNESS is enabled, then it's going to whine about * bogus LORs with various VM locks, so just skip to the * fatal trap handling directly. */ if (td->td_critnest != 0 || WITNESS_CHECK(WARN_SLEEPOK | WARN_GIANTOK, NULL, "Kernel page fault") != 0) { trap_fatal(frame, eva); return (-1); } } va = trunc_page(eva); if (va >= VM_MIN_KERNEL_ADDRESS) { /* * Don't allow user-mode faults in kernel address space. */ if (usermode) goto nogo; map = kernel_map; } else { /* * This is a fault on non-kernel virtual memory. If either * p or p->p_vmspace is NULL, then the fault is fatal. */ if (p == NULL || (vm = p->p_vmspace) == NULL) goto nogo; map = &vm->vm_map; /* * When accessing a usermode address, kernel must be * ready to accept the page fault, and provide a * handling routine. Since accessing the address * without the handler is a bug, do not try to handle * it normally, and panic immediately. */ if (!usermode && (td->td_intr_nesting_level != 0 || curpcb->pcb_onfault == NULL)) { trap_fatal(frame, eva); return (-1); } } /* * If the trap was caused by errant bits in the PTE then panic. */ if (frame->tf_err & PGEX_RSV) { trap_fatal(frame, eva); return (-1); } /* * PGEX_I is defined only if the execute disable bit capability is * supported and enabled. */ if (frame->tf_err & PGEX_W) ftype = VM_PROT_WRITE; else if ((frame->tf_err & PGEX_I) && pg_nx != 0) ftype = VM_PROT_EXECUTE; else ftype = VM_PROT_READ; if (map != kernel_map) { /* * Keep swapout from messing with us during this * critical time. */ PROC_LOCK(p); ++p->p_lock; PROC_UNLOCK(p); /* Fault in the user page: */ rv = vm_fault(map, va, ftype, VM_FAULT_NORMAL); PROC_LOCK(p); --p->p_lock; PROC_UNLOCK(p); } else { /* * Don't have to worry about process locking or stacks in the * kernel. */ rv = vm_fault(map, va, ftype, VM_FAULT_NORMAL); } if (rv == KERN_SUCCESS) { #ifdef HWPMC_HOOKS if (ftype == VM_PROT_READ || ftype == VM_PROT_WRITE) { PMC_SOFT_CALL_TF( , , page_fault, all, frame); if (ftype == VM_PROT_READ) PMC_SOFT_CALL_TF( , , page_fault, read, frame); else PMC_SOFT_CALL_TF( , , page_fault, write, frame); } #endif return (0); } nogo: if (!usermode) { if (td->td_intr_nesting_level == 0 && curpcb->pcb_onfault != NULL) { frame->tf_rip = (long)curpcb->pcb_onfault; return (0); } trap_fatal(frame, eva); return (-1); } return ((rv == KERN_PROTECTION_FAILURE) ? SIGBUS : SIGSEGV); } static void trap_fatal(frame, eva) struct trapframe *frame; vm_offset_t eva; { int code, ss; u_int type; long esp; struct soft_segment_descriptor softseg; char *msg; code = frame->tf_err; type = frame->tf_trapno; sdtossd(&gdt[NGDT * PCPU_GET(cpuid) + IDXSEL(frame->tf_cs & 0xffff)], &softseg); if (type <= MAX_TRAP_MSG) msg = trap_msg[type]; else msg = "UNKNOWN"; printf("\n\nFatal trap %d: %s while in %s mode\n", type, msg, ISPL(frame->tf_cs) == SEL_UPL ? "user" : "kernel"); #ifdef SMP /* two separate prints in case of a trap on an unmapped page */ printf("cpuid = %d; ", PCPU_GET(cpuid)); printf("apic id = %02x\n", PCPU_GET(apic_id)); #endif if (type == T_PAGEFLT) { printf("fault virtual address = 0x%lx\n", eva); printf("fault code = %s %s %s%s, %s\n", code & PGEX_U ? "user" : "supervisor", code & PGEX_W ? "write" : "read", code & PGEX_I ? "instruction" : "data", code & PGEX_RSV ? " rsv" : "", code & PGEX_P ? "protection violation" : "page not present"); } printf("instruction pointer = 0x%lx:0x%lx\n", frame->tf_cs & 0xffff, frame->tf_rip); if (ISPL(frame->tf_cs) == SEL_UPL) { ss = frame->tf_ss & 0xffff; esp = frame->tf_rsp; } else { ss = GSEL(GDATA_SEL, SEL_KPL); esp = (long)&frame->tf_rsp; } printf("stack pointer = 0x%x:0x%lx\n", ss, esp); printf("frame pointer = 0x%x:0x%lx\n", ss, frame->tf_rbp); printf("code segment = base 0x%lx, limit 0x%lx, type 0x%x\n", softseg.ssd_base, softseg.ssd_limit, softseg.ssd_type); printf(" = DPL %d, pres %d, long %d, def32 %d, gran %d\n", softseg.ssd_dpl, softseg.ssd_p, softseg.ssd_long, softseg.ssd_def32, softseg.ssd_gran); printf("processor eflags = "); if (frame->tf_rflags & PSL_T) printf("trace trap, "); if (frame->tf_rflags & PSL_I) printf("interrupt enabled, "); if (frame->tf_rflags & PSL_NT) printf("nested task, "); if (frame->tf_rflags & PSL_RF) printf("resume, "); printf("IOPL = %ld\n", (frame->tf_rflags & PSL_IOPL) >> 12); - printf("current process = "); - if (curproc) { - printf("%lu (%s)\n", - (u_long)curproc->p_pid, curthread->td_name ? - curthread->td_name : ""); - } else { - printf("Idle\n"); - } + printf("current process = %d (%s)\n", + curproc->p_pid, curthread->td_name); #ifdef KDB if (debugger_on_panic || kdb_active) if (kdb_trap(type, 0, frame)) return; #endif printf("trap number = %d\n", type); if (type <= MAX_TRAP_MSG) panic("%s", trap_msg[type]); else panic("unknown/reserved trap"); } /* * Double fault handler. Called when a fault occurs while writing * a frame for a trap/exception onto the stack. This usually occurs * when the stack overflows (such is the case with infinite recursion, * for example). */ void dblfault_handler(struct trapframe *frame) { #ifdef KDTRACE_HOOKS if (dtrace_doubletrap_func != NULL) (*dtrace_doubletrap_func)(); #endif printf("\nFatal double fault\n"); printf("rip = 0x%lx\n", frame->tf_rip); printf("rsp = 0x%lx\n", frame->tf_rsp); printf("rbp = 0x%lx\n", frame->tf_rbp); #ifdef SMP /* two separate prints in case of a trap on an unmapped page */ printf("cpuid = %d; ", PCPU_GET(cpuid)); printf("apic id = %02x\n", PCPU_GET(apic_id)); #endif panic("double fault"); } int cpu_fetch_syscall_args(struct thread *td, struct syscall_args *sa) { struct proc *p; struct trapframe *frame; register_t *argp; caddr_t params; int reg, regcnt, error; p = td->td_proc; frame = td->td_frame; reg = 0; regcnt = 6; params = (caddr_t)frame->tf_rsp + sizeof(register_t); sa->code = frame->tf_rax; if (sa->code == SYS_syscall || sa->code == SYS___syscall) { sa->code = frame->tf_rdi; reg++; regcnt--; } if (p->p_sysent->sv_mask) sa->code &= p->p_sysent->sv_mask; if (sa->code >= p->p_sysent->sv_size) sa->callp = &p->p_sysent->sv_table[0]; else sa->callp = &p->p_sysent->sv_table[sa->code]; sa->narg = sa->callp->sy_narg; KASSERT(sa->narg <= sizeof(sa->args) / sizeof(sa->args[0]), ("Too many syscall arguments!")); error = 0; argp = &frame->tf_rdi; argp += reg; bcopy(argp, sa->args, sizeof(sa->args[0]) * regcnt); if (sa->narg > regcnt) { KASSERT(params != NULL, ("copyin args with no params!")); error = copyin(params, &sa->args[regcnt], (sa->narg - regcnt) * sizeof(sa->args[0])); } if (error == 0) { td->td_retval[0] = 0; td->td_retval[1] = frame->tf_rdx; } return (error); } #include "../../kern/subr_syscall.c" /* * System call handler for native binaries. The trap frame is already * set up by the assembler trampoline and a pointer to it is saved in * td_frame. */ void amd64_syscall(struct thread *td, int traced) { struct syscall_args sa; int error; ksiginfo_t ksi; #ifdef DIAGNOSTIC if (ISPL(td->td_frame->tf_cs) != SEL_UPL) { panic("syscall"); /* NOT REACHED */ } #endif error = syscallenter(td, &sa); /* * Traced syscall. */ if (__predict_false(traced)) { td->td_frame->tf_rflags &= ~PSL_T; ksiginfo_init_trap(&ksi); ksi.ksi_signo = SIGTRAP; ksi.ksi_code = TRAP_TRACE; ksi.ksi_addr = (void *)td->td_frame->tf_rip; trapsignal(td, &ksi); } KASSERT(PCB_USER_FPU(td->td_pcb), ("System call %s returing with kernel FPU ctx leaked", syscallname(td->td_proc, sa.code))); KASSERT(td->td_pcb->pcb_save == get_pcb_user_save_td(td), ("System call %s returning with mangled pcb_save", syscallname(td->td_proc, sa.code))); syscallret(td, error, &sa); /* * If the user-supplied value of %rip is not a canonical * address, then some CPUs will trigger a ring 0 #GP during * the sysret instruction. However, the fault handler would * execute in ring 0 with the user's %gs and %rsp which would * not be safe. Instead, use the full return path which * catches the problem safely. */ if (td->td_frame->tf_rip >= VM_MAXUSER_ADDRESS) set_pcb_flags(td->td_pcb, PCB_FULL_IRET); } Index: projects/clang-trunk/sys/arm/arm/copystr.S =================================================================== --- projects/clang-trunk/sys/arm/arm/copystr.S (revision 283870) +++ projects/clang-trunk/sys/arm/arm/copystr.S (revision 283871) @@ -1,213 +1,214 @@ /* $NetBSD: copystr.S,v 1.8 2002/10/13 14:54:48 bjh21 Exp $ */ /*- * Copyright (c) 1995 Mark Brinicombe. * All rights reserved. * * Redistribution and use in source and binary forms, with or without * modification, are permitted provided that the following conditions * are met: * 1. Redistributions of source code must retain the above copyright * notice, this list of conditions and the following disclaimer. * 2. Redistributions in binary form must reproduce the above copyright * notice, this list of conditions and the following disclaimer in the * documentation and/or other materials provided with the distribution. * 3. All advertising materials mentioning features or use of this software * must display the following acknowledgement: * This product includes software developed by Mark Brinicombe. * 4. The name of the company nor the name of the author may be used to * endorse or promote products derived from this software without specific * prior written permission. * * THIS SOFTWARE IS PROVIDED BY THE AUTHOR ``AS IS'' AND ANY EXPRESS OR IMPLIED * WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF * MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. * IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, * INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES * (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR * SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF * SUCH DAMAGE. * * copystr.S * * optimised and fault protected copystr functions * * Created : 16/05/95 */ #include "assym.s" #include #include __FBSDID("$FreeBSD$"); #include .text .align 2 #ifdef _ARM_ARCH_6 #define GET_PCB(tmp) \ mrc p15, 0, tmp, c13, c0, 4; \ add tmp, tmp, #(TD_PCB) #else .Lpcb: .word _C_LABEL(__pcpu) + PC_CURPCB #define GET_PCB(tmp) \ ldr tmp, .Lpcb #endif /* * r0 - from * r1 - to * r2 - maxlens * r3 - lencopied * * Copy string from r0 to r1 */ ENTRY(copystr) stmfd sp!, {r4-r5} /* stack is 8 byte aligned */ teq r2, #0x00000000 mov r5, #0x00000000 moveq r0, #ENAMETOOLONG beq 2f 1: ldrb r4, [r0], #0x0001 add r5, r5, #0x00000001 teq r4, #0x00000000 strb r4, [r1], #0x0001 teqne r5, r2 bne 1b teq r4, #0x00000000 moveq r0, #0x00000000 movne r0, #ENAMETOOLONG 2: teq r3, #0x00000000 strne r5, [r3] ldmfd sp!, {r4-r5} /* stack is 8 byte aligned */ RET END(copystr) #define SAVE_REGS stmfd sp!, {r4-r6} #define RESTORE_REGS ldmfd sp!, {r4-r6} /* * r0 - user space address * r1 - kernel space address * r2 - maxlens * r3 - lencopied * * Copy string from user space to kernel space */ ENTRY(copyinstr) SAVE_REGS teq r2, #0x00000000 mov r6, #0x00000000 moveq r0, #ENAMETOOLONG beq 2f GET_PCB(r4) ldr r4, [r4] #ifdef DIAGNOSTIC teq r4, #0x00000000 beq .Lcopystrpcbfault #endif adr r5, .Lcopystrfault str r5, [r4, #PCB_ONFAULT] 1: ldrbt r5, [r0], #0x0001 add r6, r6, #0x00000001 teq r5, #0x00000000 strb r5, [r1], #0x0001 teqne r6, r2 bne 1b mov r0, #0x00000000 str r0, [r4, #PCB_ONFAULT] teq r5, #0x00000000 moveq r0, #0x00000000 movne r0, #ENAMETOOLONG 2: teq r3, #0x00000000 strne r6, [r3] RESTORE_REGS RET END(copyinstr) /* * r0 - kernel space address * r1 - user space address * r2 - maxlens * r3 - lencopied * * Copy string from kernel space to user space */ ENTRY(copyoutstr) SAVE_REGS teq r2, #0x00000000 mov r6, #0x00000000 moveq r0, #ENAMETOOLONG beq 2f GET_PCB(r4) ldr r4, [r4] #ifdef DIAGNOSTIC teq r4, #0x00000000 beq .Lcopystrpcbfault #endif adr r5, .Lcopystrfault str r5, [r4, #PCB_ONFAULT] 1: ldrb r5, [r0], #0x0001 add r6, r6, #0x00000001 teq r5, #0x00000000 strbt r5, [r1], #0x0001 teqne r6, r2 bne 1b mov r0, #0x00000000 str r0, [r4, #PCB_ONFAULT] teq r5, #0x00000000 moveq r0, #0x00000000 movne r0, #ENAMETOOLONG 2: teq r3, #0x00000000 strne r6, [r3] RESTORE_REGS RET END(copyoutstr) /* A fault occurred during the copy */ .Lcopystrfault: + mov r0, #EFAULT mov r1, #0x00000000 str r1, [r4, #PCB_ONFAULT] RESTORE_REGS RET #ifdef DIAGNOSTIC .Lcopystrpcbfault: mov r2, r1 mov r1, r0 adr r0, Lcopystrpcbfaulttext bic sp, sp, #7 /* align stack to 8 bytes */ b _C_LABEL(panic) Lcopystrpcbfaulttext: .asciz "No valid PCB during copyinoutstr() addr1=%08x addr2=%08x\n" .align 2 #endif Index: projects/clang-trunk/sys/arm/include/_types.h =================================================================== --- projects/clang-trunk/sys/arm/include/_types.h (revision 283870) +++ projects/clang-trunk/sys/arm/include/_types.h (revision 283871) @@ -1,132 +1,126 @@ /*- * Copyright (c) 2002 Mike Barcroft * Copyright (c) 1990, 1993 * The Regents of the University of California. All rights reserved. * * Redistribution and use in source and binary forms, with or without * modification, are permitted provided that the following conditions * are met: * 1. Redistributions of source code must retain the above copyright * notice, this list of conditions and the following disclaimer. * 2. Redistributions in binary form must reproduce the above copyright * notice, this list of conditions and the following disclaimer in the * documentation and/or other materials provided with the distribution. * 3. All advertising materials mentioning features or use of this software * must display the following acknowledgement: * This product includes software developed by the University of * California, Berkeley and its contributors. * 4. Neither the name of the University nor the names of its contributors * may be used to endorse or promote products derived from this software * without specific prior written permission. * * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF * SUCH DAMAGE. * * From: @(#)ansi.h 8.2 (Berkeley) 1/4/94 * From: @(#)types.h 8.3 (Berkeley) 1/5/94 * $FreeBSD$ */ #ifndef _MACHINE__TYPES_H_ #define _MACHINE__TYPES_H_ #ifndef _SYS_CDEFS_H_ #error this file needs sys/cdefs.h as a prerequisite #endif /* * Basic types upon which most other types are built. */ typedef signed char __int8_t; typedef unsigned char __uint8_t; typedef short __int16_t; typedef unsigned short __uint16_t; typedef int __int32_t; typedef unsigned int __uint32_t; #ifndef lint __extension__ #endif /* LONGLONG */ typedef long long __int64_t; #ifndef lint __extension__ #endif /* LONGLONG */ typedef unsigned long long __uint64_t; /* * Standard type definitions. */ typedef __uint32_t __clock_t; /* clock()... */ typedef __int32_t __critical_t; typedef double __double_t; typedef float __float_t; typedef __int32_t __intfptr_t; typedef __int64_t __intmax_t; typedef __int32_t __intptr_t; typedef __int32_t __int_fast8_t; typedef __int32_t __int_fast16_t; typedef __int32_t __int_fast32_t; typedef __int64_t __int_fast64_t; typedef __int8_t __int_least8_t; typedef __int16_t __int_least16_t; typedef __int32_t __int_least32_t; typedef __int64_t __int_least64_t; typedef __int32_t __ptrdiff_t; /* ptr1 - ptr2 */ typedef __int32_t __register_t; typedef __int32_t __segsz_t; /* segment size (in pages) */ typedef __uint32_t __size_t; /* sizeof() */ typedef __int32_t __ssize_t; /* byte count or error */ typedef __int64_t __time_t; /* time()... */ typedef __uint32_t __uintfptr_t; typedef __uint64_t __uintmax_t; typedef __uint32_t __uintptr_t; typedef __uint32_t __uint_fast8_t; typedef __uint32_t __uint_fast16_t; typedef __uint32_t __uint_fast32_t; typedef __uint64_t __uint_fast64_t; typedef __uint8_t __uint_least8_t; typedef __uint16_t __uint_least16_t; typedef __uint32_t __uint_least32_t; typedef __uint64_t __uint_least64_t; typedef __uint32_t __u_register_t; typedef __uint32_t __vm_offset_t; typedef __int64_t __vm_ooffset_t; typedef __uint32_t __vm_paddr_t; typedef __uint64_t __vm_pindex_t; typedef __uint32_t __vm_size_t; -#ifdef __ARM_EABI__ typedef unsigned int ___wchar_t; #define __WCHAR_MIN 0 /* min value for a wchar_t */ #define __WCHAR_MAX __UINT_MAX /* max value for a wchar_t */ -#else -typedef int ___wchar_t; -#define __WCHAR_MIN __INT_MIN /* min value for a wchar_t */ -#define __WCHAR_MAX __INT_MAX /* max value for a wchar_t */ -#endif /* * Unusual type definitions. */ #ifdef __GNUCLIKE_BUILTIN_VARARGS typedef __builtin_va_list __va_list; /* internally known to gcc */ #else typedef char * __va_list; #endif /* __GNUCLIKE_BUILTIN_VARARGS */ #if defined(__GNUCLIKE_BUILTIN_VAALIST) && !defined(__GNUC_VA_LIST) \ && !defined(__NO_GNUC_VA_LIST) #define __GNUC_VA_LIST typedef __va_list __gnuc_va_list; /* compatibility w/GNU headers*/ #endif #endif /* !_MACHINE__TYPES_H_ */ Index: projects/clang-trunk/sys/arm/include/asm.h =================================================================== --- projects/clang-trunk/sys/arm/include/asm.h (revision 283870) +++ projects/clang-trunk/sys/arm/include/asm.h (revision 283871) @@ -1,249 +1,249 @@ /* $NetBSD: asm.h,v 1.5 2003/08/07 16:26:53 agc Exp $ */ /*- * Copyright (c) 1990 The Regents of the University of California. * All rights reserved. * * This code is derived from software contributed to Berkeley by * William Jolitz. * * Redistribution and use in source and binary forms, with or without * modification, are permitted provided that the following conditions * are met: * 1. Redistributions of source code must retain the above copyright * notice, this list of conditions and the following disclaimer. * 2. Redistributions in binary form must reproduce the above copyright * notice, this list of conditions and the following disclaimer in the * documentation and/or other materials provided with the distribution. * 3. Neither the name of the University nor the names of its contributors * may be used to endorse or promote products derived from this software * without specific prior written permission. * * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF * SUCH DAMAGE. * * from: @(#)asm.h 5.5 (Berkeley) 5/7/91 * * $FreeBSD$ */ #ifndef _MACHINE_ASM_H_ #define _MACHINE_ASM_H_ #include #include #include #define _C_LABEL(x) x #define _ASM_LABEL(x) x #ifndef _ALIGN_TEXT # define _ALIGN_TEXT .align 2 #endif -#if defined(__ARM_EABI__) && !defined(_STANDALONE) +#ifndef _STANDALONE #define STOP_UNWINDING .cantunwind #define _FNSTART .fnstart #define _FNEND .fnend #else #define STOP_UNWINDING #define _FNSTART #define _FNEND #endif /* * gas/arm uses @ as a single comment character and thus cannot be used here. * It recognises the # instead of an @ symbol in .type directives. */ #define _ASM_TYPE_FUNCTION #function #define _ASM_TYPE_OBJECT #object /* XXX Is this still the right prologue for profiling? */ #ifdef GPROF #define _PROF_PROLOGUE \ mov ip, lr; \ bl __mcount #else #define _PROF_PROLOGUE #endif /* * EENTRY()/EEND() mark "extra" entry/exit points from a function. * LEENTRY()/LEEND() are the the same for local symbols. * The unwind info cannot handle the concept of a nested function, or a function * with multiple .fnstart directives, but some of our assembler code is written * with multiple labels to allow entry at several points. The EENTRY() macro * defines such an extra entry point without a new .fnstart, so that it's * basically just a label that you can jump to. The EEND() macro does nothing * at all, except document the exit point associated with the same-named entry. */ #define GLOBAL(x) .global x #ifdef __thumb__ #define _FUNC_MODE .code 16; .thumb_func #else #define _FUNC_MODE .code 32 #endif #define _LEENTRY(x) .type x,_ASM_TYPE_FUNCTION; _FUNC_MODE; x: #define _LEEND(x) /* nothing */ #define _EENTRY(x) GLOBAL(x); _LEENTRY(x) #define _EEND(x) _LEEND(x) #define _LENTRY(x) .text; _ALIGN_TEXT; _LEENTRY(x); _FNSTART #define _LEND(x) .size x, . - x; _FNEND #define _ENTRY(x) .text; _ALIGN_TEXT; _EENTRY(x); _FNSTART #define _END(x) _LEND(x) #define ENTRY(y) _ENTRY(_C_LABEL(y)); _PROF_PROLOGUE #define EENTRY(y) _EENTRY(_C_LABEL(y)); #define ENTRY_NP(y) _ENTRY(_C_LABEL(y)) #define EENTRY_NP(y) _EENTRY(_C_LABEL(y)) #define END(y) _END(_C_LABEL(y)) #define EEND(y) _EEND(_C_LABEL(y)) #define ASENTRY(y) _ENTRY(_ASM_LABEL(y)); _PROF_PROLOGUE #define ASLENTRY(y) _LENTRY(_ASM_LABEL(y)); _PROF_PROLOGUE #define ASEENTRY(y) _EENTRY(_ASM_LABEL(y)); #define ASLEENTRY(y) _LEENTRY(_ASM_LABEL(y)); #define ASENTRY_NP(y) _ENTRY(_ASM_LABEL(y)) #define ASLENTRY_NP(y) _LENTRY(_ASM_LABEL(y)) #define ASEENTRY_NP(y) _EENTRY(_ASM_LABEL(y)) #define ASLEENTRY_NP(y) _LEENTRY(_ASM_LABEL(y)) #define ASEND(y) _END(_ASM_LABEL(y)) #define ASLEND(y) _LEND(_ASM_LABEL(y)) #define ASEEND(y) _EEND(_ASM_LABEL(y)) #define ASLEEND(y) _LEEND(_ASM_LABEL(y)) #define ASMSTR .asciz #if defined(PIC) #define PLT_SYM(x) PIC_SYM(x, PLT) #define GOT_SYM(x) PIC_SYM(x, GOT) #define GOT_GET(x,got,sym) \ ldr x, sym; \ ldr x, [x, got] #define GOT_INIT(got,gotsym,pclabel) \ ldr got, gotsym; \ pclabel: add got, pc #ifdef __thumb__ #define GOT_INITSYM(gotsym,pclabel) \ .align 2; \ gotsym: .word _C_LABEL(_GLOBAL_OFFSET_TABLE_) - (pclabel+4) #else #define GOT_INITSYM(gotsym,pclabel) \ .align 2; \ gotsym: .word _C_LABEL(_GLOBAL_OFFSET_TABLE_) - (pclabel+8) #endif #ifdef __STDC__ #define PIC_SYM(x,y) x ## ( ## y ## ) #else #define PIC_SYM(x,y) x/**/(/**/y/**/) #endif #else #define PLT_SYM(x) x #define GOT_SYM(x) x #define GOT_GET(x,got,sym) \ ldr x, sym; #define GOT_INIT(got,gotsym,pclabel) #define GOT_INITSYM(gotsym,pclabel) #define PIC_SYM(x,y) x #endif /* PIC */ #undef __FBSDID #if !defined(lint) && !defined(STRIP_FBSDID) #define __FBSDID(s) .ident s #else #define __FBSDID(s) /* nothing */ #endif #define WEAK_ALIAS(alias,sym) \ .weak alias; \ alias = sym #ifdef __STDC__ #define WARN_REFERENCES(sym,msg) \ .stabs msg ## ,30,0,0,0 ; \ .stabs __STRING(_C_LABEL(sym)) ## ,1,0,0,0 #else #define WARN_REFERENCES(sym,msg) \ .stabs msg,30,0,0,0 ; \ .stabs __STRING(sym),1,0,0,0 #endif /* __STDC__ */ /* Exactly one of the __ARM_ARCH_*__ macros will be defined by the compiler. */ /* The _ARM_ARCH_* macros are deprecated and will be removed soon. */ /* This should be moved into another header so it can be used in * both asm and C code. machine/asm.h cannot be included in C code. */ #if defined (__ARM_ARCH_7__) || defined (__ARM_ARCH_7A__) #define _ARM_ARCH_7 #define _HAVE_ARMv7_INSTRUCTIONS 1 #endif #if defined (_HAVE_ARMv7_INSTRUCTIONS) || defined (__ARM_ARCH_6__) || \ defined (__ARM_ARCH_6J__) || defined (__ARM_ARCH_6K__) || \ defined (__ARM_ARCH_6Z__) || defined (__ARM_ARCH_6ZK__) #define _ARM_ARCH_6 #define _HAVE_ARMv6_INSTRUCTIONS 1 #endif #if defined (_HAVE_ARMv6_INSTRUCTIONS) || defined (__ARM_ARCH_5TE__) || \ defined (__ARM_ARCH_5TEJ__) || defined (__ARM_ARCH_5E__) #define _ARM_ARCH_5E #define _HAVE_ARMv5E_INSTRUCTIONS 1 #endif #if defined (_HAVE_ARMv5E_INSTRUCTIONS) || defined (__ARM_ARCH_5__) || \ defined (__ARM_ARCH_5T__) #define _ARM_ARCH_5 #define _HAVE_ARMv5_INSTRUCTIONS 1 #endif #if defined (_HAVE_ARMv5_INSTRUCTIONS) || defined (__ARM_ARCH_4T__) #define _ARM_ARCH_4T #define _HAVE_ARMv4T_INSTRUCTIONS 1 #endif /* FreeBSD requires ARMv4, so this is always set. */ #define _HAVE_ARMv4_INSTRUCTIONS 1 #if defined (_HAVE_ARMv4T_INSTRUCTIONS) # define RET bx lr # define RETeq bxeq lr # define RETne bxne lr # define RETc(c) bx##c lr #else # define RET mov pc, lr # define RETeq moveq pc, lr # define RETne movne pc, lr # define RETc(c) mov##c pc, lr #endif #if __ARM_ARCH >= 7 #define ISB isb #define DSB dsb #define DMB dmb #define WFI wfi #elif __ARM_ARCH == 6 #define ISB mcr CP15_CP15ISB #define DSB mcr CP15_CP15DSB #define DMB mcr CP15_CP15DMB #define WFI mcr CP15_CP15WFI #else #define ISB mcr CP15_CP15ISB #define DSB mcr CP15_CP15DSB /* DSB and DMB are the */ #define DMB mcr CP15_CP15DSB /* same prior to v6.*/ /* No form of WFI available on v4, define nothing to get an error on use. */ #endif #endif /* !_MACHINE_ASM_H_ */ Index: projects/clang-trunk/sys/arm/include/proc.h =================================================================== --- projects/clang-trunk/sys/arm/include/proc.h (revision 283870) +++ projects/clang-trunk/sys/arm/include/proc.h (revision 283871) @@ -1,87 +1,83 @@ /*- * Copyright (c) 1991 Regents of the University of California. * All rights reserved. * * Redistribution and use in source and binary forms, with or without * modification, are permitted provided that the following conditions * are met: * 1. Redistributions of source code must retain the above copyright * notice, this list of conditions and the following disclaimer. * 2. Redistributions in binary form must reproduce the above copyright * notice, this list of conditions and the following disclaimer in the * documentation and/or other materials provided with the distribution. * 3. All advertising materials mentioning features or use of this software * must display the following acknowledgement: * This product includes software developed by the University of * California, Berkeley and its contributors. * 4. Neither the name of the University nor the names of its contributors * may be used to endorse or promote products derived from this software * without specific prior written permission. * * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF * SUCH DAMAGE. * * from: @(#)proc.h 7.1 (Berkeley) 5/15/91 * from: FreeBSD: src/sys/i386/include/proc.h,v 1.11 2001/06/29 * $FreeBSD$ */ #ifndef _MACHINE_PROC_H_ #define _MACHINE_PROC_H_ #include struct md_utrap { utrap_entry_t *ut_precise[UT_MAX]; /* must be first */ int ut_refcnt; }; struct mdthread { int md_spinlock_count; /* (k) */ register_t md_saved_cspr; /* (k) */ register_t md_spurflt_addr; /* (k) Spurious page fault address. */ int md_ptrace_instr; int md_ptrace_addr; register_t md_tp; void *md_ras_start; void *md_ras_end; }; struct mdproc { struct md_utrap *md_utrap; void *md_sigtramp; }; -#ifdef __ARM_EABI__ #define KINFO_PROC_SIZE 816 -#else -#define KINFO_PROC_SIZE 792 -#endif #define MAXARGS 8 /* * This holds the syscall state for a single system call. * As some syscall arguments may be 64-bit aligned we need to ensure the * args value is 64-bit aligned. The ABI will then ensure any 64-bit * arguments are already correctly aligned, even if they were passed in * via registers, we just need to make sure we copy them to an algned * buffer. */ struct syscall_args { u_int code; struct sysent *callp; register_t args[MAXARGS]; int narg; u_int nap; } __aligned(8); #endif /* !_MACHINE_PROC_H_ */ Index: projects/clang-trunk/sys/boot/fdt/dts/arm/sun4i-a10.dtsi =================================================================== --- projects/clang-trunk/sys/boot/fdt/dts/arm/sun4i-a10.dtsi (revision 283870) +++ projects/clang-trunk/sys/boot/fdt/dts/arm/sun4i-a10.dtsi (revision 283871) @@ -1,141 +1,140 @@ /*- * Copyright (c) 2014 Ganbold Tsagaankhuu * All rights reserved. * * Redistribution and use in source and binary forms, with or without * modification, are permitted provided that the following conditions * are met: * 1. Redistributions of source code must retain the above copyright * notice, this list of conditions and the following disclaimer. * 2. Redistributions in binary form must reproduce the above copyright * notice, this list of conditions and the following disclaimer in the * documentation and/or other materials provided with the distribution. * * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE * ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF * SUCH DAMAGE. * * $FreeBSD$ */ / { compatible = "allwinner,sun4i-a10"; #address-cells = <1>; #size-cells = <1>; interrupt-parent = <&AINTC>; aliases { soc = &SOC; }; SOC: a10 { #address-cells = <1>; #size-cells = <1>; compatible = "simple-bus"; ranges; bus-frequency = <0>; AINTC: interrupt-controller@01c20400 { compatible = "allwinner,sun4i-ic"; interrupt-controller; #address-cells = <0>; #interrupt-cells = <1>; reg = < 0x01c20400 0x400 >; }; sramc@01c00000 { compatible = "allwinner,sun4i-sramc"; #address-cells = <1>; #size-cells = <1>; reg = < 0x01c00000 0x1000 >; }; ccm@01c20000 { compatible = "allwinner,sun4i-ccm"; #address-cells = <1>; #size-cells = <1>; reg = < 0x01c20000 0x400 >; }; timer@01c20c00 { compatible = "allwinner,sun4i-timer"; reg = <0x01c20c00 0x90>; interrupts = < 22 >; interrupt-parent = <&AINTC>; clock-frequency = < 24000000 >; }; watchdog@01c20c90 { compatible = "allwinner,sun4i-wdt"; reg = <0x01c20c90 0x08>; }; GPIO: gpio@01c20800 { #gpio-cells = <3>; compatible = "allwinner,sun4i-gpio"; gpio-controller; reg =< 0x01c20800 0x400 >; interrupts = < 28 >; interrupt-parent = <&AINTC>; }; usb1: usb@01c14000 { compatible = "allwinner,usb-ehci", "usb-ehci"; reg = <0x01c14000 0x1000>; interrupts = < 39 >; interrupt-parent = <&AINTC>; }; usb2: usb@01c1c000 { compatible = "allwinner,usb-ehci", "usb-ehci"; reg = <0x01c1c000 0x1000>; interrupts = < 40 >; interrupt-parent = <&AINTC>; }; mmc0: mmc@01c0f000 { compatible = "allwinner,sun4i-a10-mmc"; reg = <0x01c0f000 0x1000>; interrupts = <32>; interrupt-parent = <&AINTC>; status = "disabled"; }; sata@01c18000 { compatible = "allwinner,sun4i-ahci"; reg = <0x01c18000 0x1000>; interrupts = <56>; interrupt-parent = <&AINTC>; }; UART0: serial@01c28000 { compatible = "ns16550"; reg = <0x01c28000 0x400>; reg-shift = <2>; interrupts = <1>; interrupt-parent = <&AINTC>; current-speed = <115200>; clock-frequency = < 24000000 >; busy-detect = <1>; - broken-txfifo = <1>; }; emac@01c0b000 { compatible = "allwinner,sun4i-emac"; reg = <0x01c0b000 0x1000>; interrupts = <55>; interrupt-parent = <&AINTC>; }; }; }; Index: projects/clang-trunk/sys/boot/fdt/dts/arm/sun7i-a20.dtsi =================================================================== --- projects/clang-trunk/sys/boot/fdt/dts/arm/sun7i-a20.dtsi (revision 283870) +++ projects/clang-trunk/sys/boot/fdt/dts/arm/sun7i-a20.dtsi (revision 283871) @@ -1,147 +1,146 @@ /*- * Copyright (c) 2014 Ganbold Tsagaankhuu * All rights reserved. * * Redistribution and use in source and binary forms, with or without * modification, are permitted provided that the following conditions * are met: * 1. Redistributions of source code must retain the above copyright * notice, this list of conditions and the following disclaimer. * 2. Redistributions in binary form must reproduce the above copyright * notice, this list of conditions and the following disclaimer in the * documentation and/or other materials provided with the distribution. * * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE * ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF * SUCH DAMAGE. * * $FreeBSD$ */ / { compatible = "allwinner,sun7i-a20"; #address-cells = <1>; #size-cells = <1>; interrupt-parent = <&GIC>; aliases { soc = &SOC; }; SOC: a20 { #address-cells = <1>; #size-cells = <1>; compatible = "simple-bus"; ranges; bus-frequency = <0>; GIC: interrupt-controller@01c81000 { compatible = "arm,gic"; reg = <0x01c81000 0x1000>, /* Distributor Registers */ <0x01c82000 0x0100>; /* CPU Interface Registers */ interrupt-controller; #interrupt-cells = <1>; }; sramc@01c00000 { compatible = "allwinner,sun4i-sramc"; #address-cells = <1>; #size-cells = <1>; reg = < 0x01c00000 0x1000 >; }; cpu-cfg@01c25c00 { compatible = "allwinner,sun7i-cpu-cfg"; #address-cells = <1>; #size-cells = <1>; reg = < 0x01c25c00 0x400 >; }; ccm@01c20000 { compatible = "allwinner,sun4i-ccm"; #address-cells = <1>; #size-cells = <1>; reg = < 0x01c20000 0x400 >; }; timer@01c20c00 { compatible = "allwinner,sun7i-timer"; reg = <0x01c20c00 0x90>; interrupts = < 22 >; interrupt-parent = <&GIC>; clock-frequency = < 24000000 >; }; watchdog@01c20c90 { compatible = "allwinner,sun4i-wdt"; reg = <0x01c20c90 0x10>; }; GPIO: gpio@01c20800 { #gpio-cells = <3>; compatible = "allwinner,sun4i-gpio"; gpio-controller; reg =< 0x01c20800 0x400 >; interrupts = < 28 >; interrupt-parent = <&GIC>; }; usb1: usb@01c14000 { compatible = "allwinner,usb-ehci", "usb-ehci"; reg = <0x01c14000 0x1000>; interrupts = < 39 >; interrupt-parent = <&GIC>; }; usb2: usb@01c1c000 { compatible = "allwinner,usb-ehci", "usb-ehci"; reg = <0x01c1c000 0x1000>; interrupts = < 40 >; interrupt-parent = <&GIC>; }; mmc0: mmc@01c0f000 { compatible = "allwinner,sun4i-a10-mmc"; reg = <0x01c0f000 0x1000>; interrupts = <32>; interrupt-parent = <&GIC>; status = "disabled"; }; sata@01c18000 { compatible = "allwinner,sun4i-a10-ahci"; reg = <0x01c18000 0x1000>; interrupts = <56>; interrupt-parent = <&GIC>; }; UART0: serial@01c28000 { compatible = "ns16550"; reg = <0x01c28000 0x400>; reg-shift = <2>; interrupts = <1>; interrupt-parent = <&GIC>; current-speed = <115200>; clock-frequency = < 24000000 >; busy-detect = <1>; - broken-txfifo = <1>; }; emac@01c0b000 { compatible = "allwinner,sun4i-emac"; reg = <0x01c0b000 0x1000>; interrupts = <55>; interrupt-parent = <&GIC>; }; }; }; Index: projects/clang-trunk/sys/boot =================================================================== --- projects/clang-trunk/sys/boot (revision 283870) +++ projects/clang-trunk/sys/boot (revision 283871) Property changes on: projects/clang-trunk/sys/boot ___________________________________________________________________ Modified: svn:mergeinfo ## -0,0 +0,1 ## Merged /head/sys/boot:r283596-283870 Index: projects/clang-trunk/sys/dev/cxgbe/t4_main.c =================================================================== --- projects/clang-trunk/sys/dev/cxgbe/t4_main.c (revision 283870) +++ projects/clang-trunk/sys/dev/cxgbe/t4_main.c (revision 283871) @@ -1,8555 +1,8539 @@ /*- * Copyright (c) 2011 Chelsio Communications, Inc. * All rights reserved. * Written by: Navdeep Parhar * * Redistribution and use in source and binary forms, with or without * modification, are permitted provided that the following conditions * are met: * 1. Redistributions of source code must retain the above copyright * notice, this list of conditions and the following disclaimer. * 2. Redistributions in binary form must reproduce the above copyright * notice, this list of conditions and the following disclaimer in the * documentation and/or other materials provided with the distribution. * * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE * ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF * SUCH DAMAGE. */ #include __FBSDID("$FreeBSD$"); #include "opt_inet.h" #include "opt_inet6.h" #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #if defined(__i386__) || defined(__amd64__) #include #include #endif #include "common/common.h" #include "common/t4_msg.h" #include "common/t4_regs.h" #include "common/t4_regs_values.h" #include "t4_ioctl.h" #include "t4_l2t.h" #include "t4_mp_ring.h" /* T4 bus driver interface */ static int t4_probe(device_t); static int t4_attach(device_t); static int t4_detach(device_t); static device_method_t t4_methods[] = { DEVMETHOD(device_probe, t4_probe), DEVMETHOD(device_attach, t4_attach), DEVMETHOD(device_detach, t4_detach), DEVMETHOD_END }; static driver_t t4_driver = { "t4nex", t4_methods, sizeof(struct adapter) }; /* T4 port (cxgbe) interface */ static int cxgbe_probe(device_t); static int cxgbe_attach(device_t); static int cxgbe_detach(device_t); static device_method_t cxgbe_methods[] = { DEVMETHOD(device_probe, cxgbe_probe), DEVMETHOD(device_attach, cxgbe_attach), DEVMETHOD(device_detach, cxgbe_detach), { 0, 0 } }; static driver_t cxgbe_driver = { "cxgbe", cxgbe_methods, sizeof(struct port_info) }; static d_ioctl_t t4_ioctl; static d_open_t t4_open; static d_close_t t4_close; static struct cdevsw t4_cdevsw = { .d_version = D_VERSION, .d_flags = 0, .d_open = t4_open, .d_close = t4_close, .d_ioctl = t4_ioctl, .d_name = "t4nex", }; /* T5 bus driver interface */ static int t5_probe(device_t); static device_method_t t5_methods[] = { DEVMETHOD(device_probe, t5_probe), DEVMETHOD(device_attach, t4_attach), DEVMETHOD(device_detach, t4_detach), DEVMETHOD_END }; static driver_t t5_driver = { "t5nex", t5_methods, sizeof(struct adapter) }; /* T5 port (cxl) interface */ static driver_t cxl_driver = { "cxl", cxgbe_methods, sizeof(struct port_info) }; static struct cdevsw t5_cdevsw = { .d_version = D_VERSION, .d_flags = 0, .d_open = t4_open, .d_close = t4_close, .d_ioctl = t4_ioctl, .d_name = "t5nex", }; /* ifnet + media interface */ static void cxgbe_init(void *); static int cxgbe_ioctl(struct ifnet *, unsigned long, caddr_t); static int cxgbe_transmit(struct ifnet *, struct mbuf *); static void cxgbe_qflush(struct ifnet *); static uint64_t cxgbe_get_counter(struct ifnet *, ift_counter); static int cxgbe_media_change(struct ifnet *); static void cxgbe_media_status(struct ifnet *, struct ifmediareq *); MALLOC_DEFINE(M_CXGBE, "cxgbe", "Chelsio T4/T5 Ethernet driver and services"); /* * Correct lock order when you need to acquire multiple locks is t4_list_lock, * then ADAPTER_LOCK, then t4_uld_list_lock. */ static struct sx t4_list_lock; SLIST_HEAD(, adapter) t4_list; #ifdef TCP_OFFLOAD static struct sx t4_uld_list_lock; SLIST_HEAD(, uld_info) t4_uld_list; #endif /* * Tunables. See tweak_tunables() too. * * Each tunable is set to a default value here if it's known at compile-time. * Otherwise it is set to -1 as an indication to tweak_tunables() that it should * provide a reasonable default when the driver is loaded. * * Tunables applicable to both T4 and T5 are under hw.cxgbe. Those specific to * T5 are under hw.cxl. */ /* * Number of queues for tx and rx, 10G and 1G, NIC and offload. */ #define NTXQ_10G 16 static int t4_ntxq10g = -1; TUNABLE_INT("hw.cxgbe.ntxq10g", &t4_ntxq10g); #define NRXQ_10G 8 static int t4_nrxq10g = -1; TUNABLE_INT("hw.cxgbe.nrxq10g", &t4_nrxq10g); #define NTXQ_1G 4 static int t4_ntxq1g = -1; TUNABLE_INT("hw.cxgbe.ntxq1g", &t4_ntxq1g); #define NRXQ_1G 2 static int t4_nrxq1g = -1; TUNABLE_INT("hw.cxgbe.nrxq1g", &t4_nrxq1g); static int t4_rsrv_noflowq = 0; TUNABLE_INT("hw.cxgbe.rsrv_noflowq", &t4_rsrv_noflowq); #ifdef TCP_OFFLOAD #define NOFLDTXQ_10G 8 static int t4_nofldtxq10g = -1; TUNABLE_INT("hw.cxgbe.nofldtxq10g", &t4_nofldtxq10g); #define NOFLDRXQ_10G 2 static int t4_nofldrxq10g = -1; TUNABLE_INT("hw.cxgbe.nofldrxq10g", &t4_nofldrxq10g); #define NOFLDTXQ_1G 2 static int t4_nofldtxq1g = -1; TUNABLE_INT("hw.cxgbe.nofldtxq1g", &t4_nofldtxq1g); #define NOFLDRXQ_1G 1 static int t4_nofldrxq1g = -1; TUNABLE_INT("hw.cxgbe.nofldrxq1g", &t4_nofldrxq1g); #endif #ifdef DEV_NETMAP #define NNMTXQ_10G 2 static int t4_nnmtxq10g = -1; TUNABLE_INT("hw.cxgbe.nnmtxq10g", &t4_nnmtxq10g); #define NNMRXQ_10G 2 static int t4_nnmrxq10g = -1; TUNABLE_INT("hw.cxgbe.nnmrxq10g", &t4_nnmrxq10g); #define NNMTXQ_1G 1 static int t4_nnmtxq1g = -1; TUNABLE_INT("hw.cxgbe.nnmtxq1g", &t4_nnmtxq1g); #define NNMRXQ_1G 1 static int t4_nnmrxq1g = -1; TUNABLE_INT("hw.cxgbe.nnmrxq1g", &t4_nnmrxq1g); #endif /* * Holdoff parameters for 10G and 1G ports. */ #define TMR_IDX_10G 1 static int t4_tmr_idx_10g = TMR_IDX_10G; TUNABLE_INT("hw.cxgbe.holdoff_timer_idx_10G", &t4_tmr_idx_10g); #define PKTC_IDX_10G (-1) static int t4_pktc_idx_10g = PKTC_IDX_10G; TUNABLE_INT("hw.cxgbe.holdoff_pktc_idx_10G", &t4_pktc_idx_10g); #define TMR_IDX_1G 1 static int t4_tmr_idx_1g = TMR_IDX_1G; TUNABLE_INT("hw.cxgbe.holdoff_timer_idx_1G", &t4_tmr_idx_1g); #define PKTC_IDX_1G (-1) static int t4_pktc_idx_1g = PKTC_IDX_1G; TUNABLE_INT("hw.cxgbe.holdoff_pktc_idx_1G", &t4_pktc_idx_1g); /* * Size (# of entries) of each tx and rx queue. */ static unsigned int t4_qsize_txq = TX_EQ_QSIZE; TUNABLE_INT("hw.cxgbe.qsize_txq", &t4_qsize_txq); static unsigned int t4_qsize_rxq = RX_IQ_QSIZE; TUNABLE_INT("hw.cxgbe.qsize_rxq", &t4_qsize_rxq); /* * Interrupt types allowed (bits 0, 1, 2 = INTx, MSI, MSI-X respectively). */ static int t4_intr_types = INTR_MSIX | INTR_MSI | INTR_INTX; TUNABLE_INT("hw.cxgbe.interrupt_types", &t4_intr_types); /* * Configuration file. */ #define DEFAULT_CF "default" #define FLASH_CF "flash" #define UWIRE_CF "uwire" #define FPGA_CF "fpga" static char t4_cfg_file[32] = DEFAULT_CF; TUNABLE_STR("hw.cxgbe.config_file", t4_cfg_file, sizeof(t4_cfg_file)); /* * PAUSE settings (bit 0, 1 = rx_pause, tx_pause respectively). * rx_pause = 1 to heed incoming PAUSE frames, 0 to ignore them. * tx_pause = 1 to emit PAUSE frames when the rx FIFO reaches its high water * mark or when signalled to do so, 0 to never emit PAUSE. */ static int t4_pause_settings = PAUSE_TX | PAUSE_RX; TUNABLE_INT("hw.cxgbe.pause_settings", &t4_pause_settings); /* * Firmware auto-install by driver during attach (0, 1, 2 = prohibited, allowed, * encouraged respectively). */ static unsigned int t4_fw_install = 1; TUNABLE_INT("hw.cxgbe.fw_install", &t4_fw_install); /* * ASIC features that will be used. Disable the ones you don't want so that the * chip resources aren't wasted on features that will not be used. */ static int t4_linkcaps_allowed = 0; /* No DCBX, PPP, etc. by default */ TUNABLE_INT("hw.cxgbe.linkcaps_allowed", &t4_linkcaps_allowed); static int t4_niccaps_allowed = FW_CAPS_CONFIG_NIC; TUNABLE_INT("hw.cxgbe.niccaps_allowed", &t4_niccaps_allowed); static int t4_toecaps_allowed = -1; TUNABLE_INT("hw.cxgbe.toecaps_allowed", &t4_toecaps_allowed); static int t4_rdmacaps_allowed = 0; TUNABLE_INT("hw.cxgbe.rdmacaps_allowed", &t4_rdmacaps_allowed); static int t4_iscsicaps_allowed = 0; TUNABLE_INT("hw.cxgbe.iscsicaps_allowed", &t4_iscsicaps_allowed); static int t4_fcoecaps_allowed = 0; TUNABLE_INT("hw.cxgbe.fcoecaps_allowed", &t4_fcoecaps_allowed); static int t5_write_combine = 0; TUNABLE_INT("hw.cxl.write_combine", &t5_write_combine); struct intrs_and_queues { uint16_t intr_type; /* INTx, MSI, or MSI-X */ uint16_t nirq; /* Total # of vectors */ uint16_t intr_flags_10g;/* Interrupt flags for each 10G port */ uint16_t intr_flags_1g; /* Interrupt flags for each 1G port */ uint16_t ntxq10g; /* # of NIC txq's for each 10G port */ uint16_t nrxq10g; /* # of NIC rxq's for each 10G port */ uint16_t ntxq1g; /* # of NIC txq's for each 1G port */ uint16_t nrxq1g; /* # of NIC rxq's for each 1G port */ uint16_t rsrv_noflowq; /* Flag whether to reserve queue 0 */ #ifdef TCP_OFFLOAD uint16_t nofldtxq10g; /* # of TOE txq's for each 10G port */ uint16_t nofldrxq10g; /* # of TOE rxq's for each 10G port */ uint16_t nofldtxq1g; /* # of TOE txq's for each 1G port */ uint16_t nofldrxq1g; /* # of TOE rxq's for each 1G port */ #endif #ifdef DEV_NETMAP uint16_t nnmtxq10g; /* # of netmap txq's for each 10G port */ uint16_t nnmrxq10g; /* # of netmap rxq's for each 10G port */ uint16_t nnmtxq1g; /* # of netmap txq's for each 1G port */ uint16_t nnmrxq1g; /* # of netmap rxq's for each 1G port */ #endif }; struct filter_entry { uint32_t valid:1; /* filter allocated and valid */ uint32_t locked:1; /* filter is administratively locked */ uint32_t pending:1; /* filter action is pending firmware reply */ uint32_t smtidx:8; /* Source MAC Table index for smac */ struct l2t_entry *l2t; /* Layer Two Table entry for dmac */ struct t4_filter_specification fs; }; static int map_bars_0_and_4(struct adapter *); static int map_bar_2(struct adapter *); static void setup_memwin(struct adapter *); static int validate_mem_range(struct adapter *, uint32_t, int); static int fwmtype_to_hwmtype(int); static int validate_mt_off_len(struct adapter *, int, uint32_t, int, uint32_t *); static void memwin_info(struct adapter *, int, uint32_t *, uint32_t *); static uint32_t position_memwin(struct adapter *, int, uint32_t); static int cfg_itype_and_nqueues(struct adapter *, int, int, struct intrs_and_queues *); static int prep_firmware(struct adapter *); static int partition_resources(struct adapter *, const struct firmware *, const char *); static int get_params__pre_init(struct adapter *); static int get_params__post_init(struct adapter *); static int set_params__post_init(struct adapter *); static void t4_set_desc(struct adapter *); static void build_medialist(struct port_info *, struct ifmedia *); static int cxgbe_init_synchronized(struct port_info *); static int cxgbe_uninit_synchronized(struct port_info *); static int setup_intr_handlers(struct adapter *); static void quiesce_txq(struct adapter *, struct sge_txq *); static void quiesce_wrq(struct adapter *, struct sge_wrq *); static void quiesce_iq(struct adapter *, struct sge_iq *); static void quiesce_fl(struct adapter *, struct sge_fl *); static int t4_alloc_irq(struct adapter *, struct irq *, int rid, driver_intr_t *, void *, char *); static int t4_free_irq(struct adapter *, struct irq *); static void reg_block_dump(struct adapter *, uint8_t *, unsigned int, unsigned int); static void t4_get_regs(struct adapter *, struct t4_regdump *, uint8_t *); static void cxgbe_refresh_stats(struct adapter *, struct port_info *); static void cxgbe_tick(void *); static void cxgbe_vlan_config(void *, struct ifnet *, uint16_t); static int cpl_not_handled(struct sge_iq *, const struct rss_header *, struct mbuf *); static int an_not_handled(struct sge_iq *, const struct rsp_ctrl *); static int fw_msg_not_handled(struct adapter *, const __be64 *); static int t4_sysctls(struct adapter *); static int cxgbe_sysctls(struct port_info *); static int sysctl_int_array(SYSCTL_HANDLER_ARGS); static int sysctl_bitfield(SYSCTL_HANDLER_ARGS); static int sysctl_btphy(SYSCTL_HANDLER_ARGS); static int sysctl_noflowq(SYSCTL_HANDLER_ARGS); static int sysctl_holdoff_tmr_idx(SYSCTL_HANDLER_ARGS); static int sysctl_holdoff_pktc_idx(SYSCTL_HANDLER_ARGS); static int sysctl_qsize_rxq(SYSCTL_HANDLER_ARGS); static int sysctl_qsize_txq(SYSCTL_HANDLER_ARGS); static int sysctl_pause_settings(SYSCTL_HANDLER_ARGS); static int sysctl_handle_t4_reg64(SYSCTL_HANDLER_ARGS); static int sysctl_temperature(SYSCTL_HANDLER_ARGS); #ifdef SBUF_DRAIN static int sysctl_cctrl(SYSCTL_HANDLER_ARGS); static int sysctl_cim_ibq_obq(SYSCTL_HANDLER_ARGS); static int sysctl_cim_la(SYSCTL_HANDLER_ARGS); static int sysctl_cim_ma_la(SYSCTL_HANDLER_ARGS); static int sysctl_cim_pif_la(SYSCTL_HANDLER_ARGS); static int sysctl_cim_qcfg(SYSCTL_HANDLER_ARGS); static int sysctl_cpl_stats(SYSCTL_HANDLER_ARGS); static int sysctl_ddp_stats(SYSCTL_HANDLER_ARGS); static int sysctl_devlog(SYSCTL_HANDLER_ARGS); static int sysctl_fcoe_stats(SYSCTL_HANDLER_ARGS); static int sysctl_hw_sched(SYSCTL_HANDLER_ARGS); static int sysctl_lb_stats(SYSCTL_HANDLER_ARGS); static int sysctl_linkdnrc(SYSCTL_HANDLER_ARGS); static int sysctl_meminfo(SYSCTL_HANDLER_ARGS); static int sysctl_mps_tcam(SYSCTL_HANDLER_ARGS); static int sysctl_path_mtus(SYSCTL_HANDLER_ARGS); static int sysctl_pm_stats(SYSCTL_HANDLER_ARGS); static int sysctl_rdma_stats(SYSCTL_HANDLER_ARGS); static int sysctl_tcp_stats(SYSCTL_HANDLER_ARGS); static int sysctl_tids(SYSCTL_HANDLER_ARGS); static int sysctl_tp_err_stats(SYSCTL_HANDLER_ARGS); static int sysctl_tp_la(SYSCTL_HANDLER_ARGS); static int sysctl_tx_rate(SYSCTL_HANDLER_ARGS); static int sysctl_ulprx_la(SYSCTL_HANDLER_ARGS); static int sysctl_wcwr_stats(SYSCTL_HANDLER_ARGS); #endif static uint32_t fconf_to_mode(uint32_t); static uint32_t mode_to_fconf(uint32_t); static uint32_t fspec_to_fconf(struct t4_filter_specification *); static int get_filter_mode(struct adapter *, uint32_t *); static int set_filter_mode(struct adapter *, uint32_t); static inline uint64_t get_filter_hits(struct adapter *, uint32_t); static int get_filter(struct adapter *, struct t4_filter *); static int set_filter(struct adapter *, struct t4_filter *); static int del_filter(struct adapter *, struct t4_filter *); static void clear_filter(struct filter_entry *); static int set_filter_wr(struct adapter *, int); static int del_filter_wr(struct adapter *, int); static int get_sge_context(struct adapter *, struct t4_sge_context *); static int load_fw(struct adapter *, struct t4_data *); static int read_card_mem(struct adapter *, int, struct t4_mem_range *); static int read_i2c(struct adapter *, struct t4_i2c_data *); static int set_sched_class(struct adapter *, struct t4_sched_params *); static int set_sched_queue(struct adapter *, struct t4_sched_queue *); #ifdef TCP_OFFLOAD static int toe_capability(struct port_info *, int); #endif static int mod_event(module_t, int, void *); struct { uint16_t device; char *desc; } t4_pciids[] = { {0xa000, "Chelsio Terminator 4 FPGA"}, {0x4400, "Chelsio T440-dbg"}, {0x4401, "Chelsio T420-CR"}, {0x4402, "Chelsio T422-CR"}, {0x4403, "Chelsio T440-CR"}, {0x4404, "Chelsio T420-BCH"}, {0x4405, "Chelsio T440-BCH"}, {0x4406, "Chelsio T440-CH"}, {0x4407, "Chelsio T420-SO"}, {0x4408, "Chelsio T420-CX"}, {0x4409, "Chelsio T420-BT"}, {0x440a, "Chelsio T404-BT"}, {0x440e, "Chelsio T440-LP-CR"}, }, t5_pciids[] = { {0xb000, "Chelsio Terminator 5 FPGA"}, {0x5400, "Chelsio T580-dbg"}, {0x5401, "Chelsio T520-CR"}, /* 2 x 10G */ {0x5402, "Chelsio T522-CR"}, /* 2 x 10G, 2 X 1G */ {0x5403, "Chelsio T540-CR"}, /* 4 x 10G */ {0x5407, "Chelsio T520-SO"}, /* 2 x 10G, nomem */ {0x5409, "Chelsio T520-BT"}, /* 2 x 10GBaseT */ {0x540a, "Chelsio T504-BT"}, /* 4 x 1G */ {0x540d, "Chelsio T580-CR"}, /* 2 x 40G */ {0x540e, "Chelsio T540-LP-CR"}, /* 4 x 10G */ {0x5410, "Chelsio T580-LP-CR"}, /* 2 x 40G */ {0x5411, "Chelsio T520-LL-CR"}, /* 2 x 10G */ {0x5412, "Chelsio T560-CR"}, /* 1 x 40G, 2 x 10G */ {0x5414, "Chelsio T580-LP-SO-CR"}, /* 2 x 40G, nomem */ {0x5415, "Chelsio T502-BT"}, /* 2 x 1G */ #ifdef notyet {0x5404, "Chelsio T520-BCH"}, {0x5405, "Chelsio T540-BCH"}, {0x5406, "Chelsio T540-CH"}, {0x5408, "Chelsio T520-CX"}, {0x540b, "Chelsio B520-SR"}, {0x540c, "Chelsio B504-BT"}, {0x540f, "Chelsio Amsterdam"}, {0x5413, "Chelsio T580-CHR"}, #endif }; #ifdef TCP_OFFLOAD /* * service_iq() has an iq and needs the fl. Offset of fl from the iq should be * exactly the same for both rxq and ofld_rxq. */ CTASSERT(offsetof(struct sge_ofld_rxq, iq) == offsetof(struct sge_rxq, iq)); CTASSERT(offsetof(struct sge_ofld_rxq, fl) == offsetof(struct sge_rxq, fl)); #endif /* No easy way to include t4_msg.h before adapter.h so we check this way */ CTASSERT(nitems(((struct adapter *)0)->cpl_handler) == NUM_CPL_CMDS); CTASSERT(nitems(((struct adapter *)0)->fw_msg_handler) == NUM_FW6_TYPES); CTASSERT(sizeof(struct cluster_metadata) <= CL_METADATA_SIZE); static int t4_probe(device_t dev) { int i; uint16_t v = pci_get_vendor(dev); uint16_t d = pci_get_device(dev); uint8_t f = pci_get_function(dev); if (v != PCI_VENDOR_ID_CHELSIO) return (ENXIO); /* Attach only to PF0 of the FPGA */ if (d == 0xa000 && f != 0) return (ENXIO); for (i = 0; i < nitems(t4_pciids); i++) { if (d == t4_pciids[i].device) { device_set_desc(dev, t4_pciids[i].desc); return (BUS_PROBE_DEFAULT); } } return (ENXIO); } static int t5_probe(device_t dev) { int i; uint16_t v = pci_get_vendor(dev); uint16_t d = pci_get_device(dev); uint8_t f = pci_get_function(dev); if (v != PCI_VENDOR_ID_CHELSIO) return (ENXIO); /* Attach only to PF0 of the FPGA */ if (d == 0xb000 && f != 0) return (ENXIO); for (i = 0; i < nitems(t5_pciids); i++) { if (d == t5_pciids[i].device) { device_set_desc(dev, t5_pciids[i].desc); return (BUS_PROBE_DEFAULT); } } return (ENXIO); } static int t4_attach(device_t dev) { struct adapter *sc; int rc = 0, i, n10g, n1g, rqidx, tqidx; struct intrs_and_queues iaq; struct sge *s; #ifdef TCP_OFFLOAD int ofld_rqidx, ofld_tqidx; #endif #ifdef DEV_NETMAP int nm_rqidx, nm_tqidx; #endif - const char *pcie_ts; sc = device_get_softc(dev); sc->dev = dev; pci_enable_busmaster(dev); if (pci_find_cap(dev, PCIY_EXPRESS, &i) == 0) { uint32_t v; pci_set_max_read_req(dev, 4096); v = pci_read_config(dev, i + PCIER_DEVICE_CTL, 2); v |= PCIEM_CTL_RELAXED_ORD_ENABLE; pci_write_config(dev, i + PCIER_DEVICE_CTL, v, 2); sc->params.pci.mps = 128 << ((v & PCIEM_CTL_MAX_PAYLOAD) >> 5); } sc->traceq = -1; mtx_init(&sc->ifp_lock, sc->ifp_lockname, 0, MTX_DEF); snprintf(sc->ifp_lockname, sizeof(sc->ifp_lockname), "%s tracer", device_get_nameunit(dev)); snprintf(sc->lockname, sizeof(sc->lockname), "%s", device_get_nameunit(dev)); mtx_init(&sc->sc_lock, sc->lockname, 0, MTX_DEF); sx_xlock(&t4_list_lock); SLIST_INSERT_HEAD(&t4_list, sc, link); sx_xunlock(&t4_list_lock); mtx_init(&sc->sfl_lock, "starving freelists", 0, MTX_DEF); TAILQ_INIT(&sc->sfl); callout_init(&sc->sfl_callout, 1); mtx_init(&sc->regwin_lock, "register and memory window", 0, MTX_DEF); rc = map_bars_0_and_4(sc); if (rc != 0) goto done; /* error message displayed already */ /* * This is the real PF# to which we're attaching. Works from within PCI * passthrough environments too, where pci_get_function() could return a * different PF# depending on the passthrough configuration. We need to * use the real PF# in all our communication with the firmware. */ sc->pf = G_SOURCEPF(t4_read_reg(sc, A_PL_WHOAMI)); sc->mbox = sc->pf; memset(sc->chan_map, 0xff, sizeof(sc->chan_map)); sc->an_handler = an_not_handled; for (i = 0; i < nitems(sc->cpl_handler); i++) sc->cpl_handler[i] = cpl_not_handled; for (i = 0; i < nitems(sc->fw_msg_handler); i++) sc->fw_msg_handler[i] = fw_msg_not_handled; t4_register_cpl_handler(sc, CPL_SET_TCB_RPL, t4_filter_rpl); t4_register_cpl_handler(sc, CPL_TRACE_PKT, t4_trace_pkt); t4_register_cpl_handler(sc, CPL_TRACE_PKT_T5, t5_trace_pkt); t4_init_sge_cpl_handlers(sc); /* Prepare the adapter for operation */ rc = -t4_prep_adapter(sc); if (rc != 0) { device_printf(dev, "failed to prepare adapter: %d.\n", rc); goto done; } /* * Do this really early, with the memory windows set up even before the * character device. The userland tool's register i/o and mem read * will work even in "recovery mode". */ setup_memwin(sc); sc->cdev = make_dev(is_t4(sc) ? &t4_cdevsw : &t5_cdevsw, device_get_unit(dev), UID_ROOT, GID_WHEEL, 0600, "%s", device_get_nameunit(dev)); if (sc->cdev == NULL) device_printf(dev, "failed to create nexus char device.\n"); else sc->cdev->si_drv1 = sc; /* Go no further if recovery mode has been requested. */ if (TUNABLE_INT_FETCH("hw.cxgbe.sos", &i) && i != 0) { device_printf(dev, "recovery mode.\n"); goto done; } #if defined(__i386__) if ((cpu_feature & CPUID_CX8) == 0) { device_printf(dev, "64 bit atomics not available.\n"); rc = ENOTSUP; goto done; } #endif /* Prepare the firmware for operation */ rc = prep_firmware(sc); if (rc != 0) goto done; /* error message displayed already */ rc = get_params__post_init(sc); if (rc != 0) goto done; /* error message displayed already */ rc = set_params__post_init(sc); if (rc != 0) goto done; /* error message displayed already */ rc = map_bar_2(sc); if (rc != 0) goto done; /* error message displayed already */ rc = t4_create_dma_tag(sc); if (rc != 0) goto done; /* error message displayed already */ /* * First pass over all the ports - allocate VIs and initialize some * basic parameters like mac address, port type, etc. We also figure * out whether a port is 10G or 1G and use that information when * calculating how many interrupts to attempt to allocate. */ n10g = n1g = 0; for_each_port(sc, i) { struct port_info *pi; pi = malloc(sizeof(*pi), M_CXGBE, M_ZERO | M_WAITOK); sc->port[i] = pi; /* These must be set before t4_port_init */ pi->adapter = sc; pi->port_id = i; /* Allocate the vi and initialize parameters like mac addr */ rc = -t4_port_init(pi, sc->mbox, sc->pf, 0); if (rc != 0) { device_printf(dev, "unable to initialize port %d: %d\n", i, rc); free(pi, M_CXGBE); sc->port[i] = NULL; goto done; } pi->link_cfg.requested_fc &= ~(PAUSE_TX | PAUSE_RX); pi->link_cfg.requested_fc |= t4_pause_settings; pi->link_cfg.fc &= ~(PAUSE_TX | PAUSE_RX); pi->link_cfg.fc |= t4_pause_settings; rc = -t4_link_start(sc, sc->mbox, pi->tx_chan, &pi->link_cfg); if (rc != 0) { device_printf(dev, "port %d l1cfg failed: %d\n", i, rc); free(pi, M_CXGBE); sc->port[i] = NULL; goto done; } snprintf(pi->lockname, sizeof(pi->lockname), "%sp%d", device_get_nameunit(dev), i); mtx_init(&pi->pi_lock, pi->lockname, 0, MTX_DEF); sc->chan_map[pi->tx_chan] = i; if (is_10G_port(pi) || is_40G_port(pi)) { n10g++; pi->tmr_idx = t4_tmr_idx_10g; pi->pktc_idx = t4_pktc_idx_10g; } else { n1g++; pi->tmr_idx = t4_tmr_idx_1g; pi->pktc_idx = t4_pktc_idx_1g; } pi->xact_addr_filt = -1; pi->linkdnrc = -1; pi->qsize_rxq = t4_qsize_rxq; pi->qsize_txq = t4_qsize_txq; pi->dev = device_add_child(dev, is_t4(sc) ? "cxgbe" : "cxl", -1); if (pi->dev == NULL) { device_printf(dev, "failed to add device for port %d.\n", i); rc = ENXIO; goto done; } device_set_softc(pi->dev, pi); } /* * Interrupt type, # of interrupts, # of rx/tx queues, etc. */ rc = cfg_itype_and_nqueues(sc, n10g, n1g, &iaq); if (rc != 0) goto done; /* error message displayed already */ sc->intr_type = iaq.intr_type; sc->intr_count = iaq.nirq; s = &sc->sge; s->nrxq = n10g * iaq.nrxq10g + n1g * iaq.nrxq1g; s->ntxq = n10g * iaq.ntxq10g + n1g * iaq.ntxq1g; s->neq = s->ntxq + s->nrxq; /* the free list in an rxq is an eq */ s->neq += sc->params.nports + 1;/* ctrl queues: 1 per port + 1 mgmt */ s->niq = s->nrxq + 1; /* 1 extra for firmware event queue */ #ifdef TCP_OFFLOAD if (is_offload(sc)) { s->nofldrxq = n10g * iaq.nofldrxq10g + n1g * iaq.nofldrxq1g; s->nofldtxq = n10g * iaq.nofldtxq10g + n1g * iaq.nofldtxq1g; s->neq += s->nofldtxq + s->nofldrxq; s->niq += s->nofldrxq; s->ofld_rxq = malloc(s->nofldrxq * sizeof(struct sge_ofld_rxq), M_CXGBE, M_ZERO | M_WAITOK); s->ofld_txq = malloc(s->nofldtxq * sizeof(struct sge_wrq), M_CXGBE, M_ZERO | M_WAITOK); } #endif #ifdef DEV_NETMAP s->nnmrxq = n10g * iaq.nnmrxq10g + n1g * iaq.nnmrxq1g; s->nnmtxq = n10g * iaq.nnmtxq10g + n1g * iaq.nnmtxq1g; s->neq += s->nnmtxq + s->nnmrxq; s->niq += s->nnmrxq; s->nm_rxq = malloc(s->nnmrxq * sizeof(struct sge_nm_rxq), M_CXGBE, M_ZERO | M_WAITOK); s->nm_txq = malloc(s->nnmtxq * sizeof(struct sge_nm_txq), M_CXGBE, M_ZERO | M_WAITOK); #endif s->ctrlq = malloc(sc->params.nports * sizeof(struct sge_wrq), M_CXGBE, M_ZERO | M_WAITOK); s->rxq = malloc(s->nrxq * sizeof(struct sge_rxq), M_CXGBE, M_ZERO | M_WAITOK); s->txq = malloc(s->ntxq * sizeof(struct sge_txq), M_CXGBE, M_ZERO | M_WAITOK); s->iqmap = malloc(s->niq * sizeof(struct sge_iq *), M_CXGBE, M_ZERO | M_WAITOK); s->eqmap = malloc(s->neq * sizeof(struct sge_eq *), M_CXGBE, M_ZERO | M_WAITOK); sc->irq = malloc(sc->intr_count * sizeof(struct irq), M_CXGBE, M_ZERO | M_WAITOK); t4_init_l2t(sc, M_WAITOK); /* * Second pass over the ports. This time we know the number of rx and * tx queues that each port should get. */ rqidx = tqidx = 0; #ifdef TCP_OFFLOAD ofld_rqidx = ofld_tqidx = 0; #endif #ifdef DEV_NETMAP nm_rqidx = nm_tqidx = 0; #endif for_each_port(sc, i) { struct port_info *pi = sc->port[i]; if (pi == NULL) continue; pi->first_rxq = rqidx; pi->first_txq = tqidx; if (is_10G_port(pi) || is_40G_port(pi)) { pi->flags |= iaq.intr_flags_10g; pi->nrxq = iaq.nrxq10g; pi->ntxq = iaq.ntxq10g; } else { pi->flags |= iaq.intr_flags_1g; pi->nrxq = iaq.nrxq1g; pi->ntxq = iaq.ntxq1g; } if (pi->ntxq > 1) pi->rsrv_noflowq = iaq.rsrv_noflowq ? 1 : 0; else pi->rsrv_noflowq = 0; rqidx += pi->nrxq; tqidx += pi->ntxq; #ifdef TCP_OFFLOAD if (is_offload(sc)) { pi->first_ofld_rxq = ofld_rqidx; pi->first_ofld_txq = ofld_tqidx; if (is_10G_port(pi) || is_40G_port(pi)) { pi->nofldrxq = iaq.nofldrxq10g; pi->nofldtxq = iaq.nofldtxq10g; } else { pi->nofldrxq = iaq.nofldrxq1g; pi->nofldtxq = iaq.nofldtxq1g; } ofld_rqidx += pi->nofldrxq; ofld_tqidx += pi->nofldtxq; } #endif #ifdef DEV_NETMAP pi->first_nm_rxq = nm_rqidx; pi->first_nm_txq = nm_tqidx; if (is_10G_port(pi) || is_40G_port(pi)) { pi->nnmrxq = iaq.nnmrxq10g; pi->nnmtxq = iaq.nnmtxq10g; } else { pi->nnmrxq = iaq.nnmrxq1g; pi->nnmtxq = iaq.nnmtxq1g; } nm_rqidx += pi->nnmrxq; nm_tqidx += pi->nnmtxq; #endif } rc = setup_intr_handlers(sc); if (rc != 0) { device_printf(dev, "failed to setup interrupt handlers: %d\n", rc); goto done; } rc = bus_generic_attach(dev); if (rc != 0) { device_printf(dev, "failed to attach all child ports: %d\n", rc); goto done; } - switch (sc->params.pci.speed) { - case 0x1: - pcie_ts = "2.5"; - break; - case 0x2: - pcie_ts = "5.0"; - break; - case 0x3: - pcie_ts = "8.0"; - break; - default: - pcie_ts = "??"; - break; - } device_printf(dev, - "PCIe x%d (%s GTS/s) (%d), %d ports, %d %s interrupt%s, %d eq, %d iq\n", - sc->params.pci.width, pcie_ts, sc->params.pci.speed, - sc->params.nports, sc->intr_count, - sc->intr_type == INTR_MSIX ? "MSI-X" : + "PCIe gen%d x%d, %d ports, %d %s interrupt%s, %d eq, %d iq\n", + sc->params.pci.speed, sc->params.pci.width, sc->params.nports, + sc->intr_count, sc->intr_type == INTR_MSIX ? "MSI-X" : (sc->intr_type == INTR_MSI ? "MSI" : "INTx"), sc->intr_count > 1 ? "s" : "", sc->sge.neq, sc->sge.niq); t4_set_desc(sc); done: if (rc != 0 && sc->cdev) { /* cdev was created and so cxgbetool works; recover that way. */ device_printf(dev, "error during attach, adapter is now in recovery mode.\n"); rc = 0; } if (rc != 0) t4_detach(dev); else t4_sysctls(sc); return (rc); } /* * Idempotent */ static int t4_detach(device_t dev) { struct adapter *sc; struct port_info *pi; int i, rc; sc = device_get_softc(dev); if (sc->flags & FULL_INIT_DONE) t4_intr_disable(sc); if (sc->cdev) { destroy_dev(sc->cdev); sc->cdev = NULL; } rc = bus_generic_detach(dev); if (rc) { device_printf(dev, "failed to detach child devices: %d\n", rc); return (rc); } for (i = 0; i < sc->intr_count; i++) t4_free_irq(sc, &sc->irq[i]); for (i = 0; i < MAX_NPORTS; i++) { pi = sc->port[i]; if (pi) { t4_free_vi(sc, sc->mbox, sc->pf, 0, pi->viid); if (pi->dev) device_delete_child(dev, pi->dev); mtx_destroy(&pi->pi_lock); free(pi, M_CXGBE); } } if (sc->flags & FULL_INIT_DONE) adapter_full_uninit(sc); if (sc->flags & FW_OK) t4_fw_bye(sc, sc->mbox); if (sc->intr_type == INTR_MSI || sc->intr_type == INTR_MSIX) pci_release_msi(dev); if (sc->regs_res) bus_release_resource(dev, SYS_RES_MEMORY, sc->regs_rid, sc->regs_res); if (sc->udbs_res) bus_release_resource(dev, SYS_RES_MEMORY, sc->udbs_rid, sc->udbs_res); if (sc->msix_res) bus_release_resource(dev, SYS_RES_MEMORY, sc->msix_rid, sc->msix_res); if (sc->l2t) t4_free_l2t(sc->l2t); #ifdef TCP_OFFLOAD free(sc->sge.ofld_rxq, M_CXGBE); free(sc->sge.ofld_txq, M_CXGBE); #endif #ifdef DEV_NETMAP free(sc->sge.nm_rxq, M_CXGBE); free(sc->sge.nm_txq, M_CXGBE); #endif free(sc->irq, M_CXGBE); free(sc->sge.rxq, M_CXGBE); free(sc->sge.txq, M_CXGBE); free(sc->sge.ctrlq, M_CXGBE); free(sc->sge.iqmap, M_CXGBE); free(sc->sge.eqmap, M_CXGBE); free(sc->tids.ftid_tab, M_CXGBE); t4_destroy_dma_tag(sc); if (mtx_initialized(&sc->sc_lock)) { sx_xlock(&t4_list_lock); SLIST_REMOVE(&t4_list, sc, adapter, link); sx_xunlock(&t4_list_lock); mtx_destroy(&sc->sc_lock); } if (mtx_initialized(&sc->tids.ftid_lock)) mtx_destroy(&sc->tids.ftid_lock); if (mtx_initialized(&sc->sfl_lock)) mtx_destroy(&sc->sfl_lock); if (mtx_initialized(&sc->ifp_lock)) mtx_destroy(&sc->ifp_lock); if (mtx_initialized(&sc->regwin_lock)) mtx_destroy(&sc->regwin_lock); bzero(sc, sizeof(*sc)); return (0); } static int cxgbe_probe(device_t dev) { char buf[128]; struct port_info *pi = device_get_softc(dev); snprintf(buf, sizeof(buf), "port %d", pi->port_id); device_set_desc_copy(dev, buf); return (BUS_PROBE_DEFAULT); } #define T4_CAP (IFCAP_VLAN_HWTAGGING | IFCAP_VLAN_MTU | IFCAP_HWCSUM | \ IFCAP_VLAN_HWCSUM | IFCAP_TSO | IFCAP_JUMBO_MTU | IFCAP_LRO | \ IFCAP_VLAN_HWTSO | IFCAP_LINKSTATE | IFCAP_HWCSUM_IPV6 | IFCAP_HWSTATS) #define T4_CAP_ENABLE (T4_CAP) static int cxgbe_attach(device_t dev) { struct port_info *pi = device_get_softc(dev); struct ifnet *ifp; char *s; int n, o; /* Allocate an ifnet and set it up */ ifp = if_alloc(IFT_ETHER); if (ifp == NULL) { device_printf(dev, "Cannot allocate ifnet\n"); return (ENOMEM); } pi->ifp = ifp; ifp->if_softc = pi; callout_init(&pi->tick, 1); if_initname(ifp, device_get_name(dev), device_get_unit(dev)); ifp->if_flags = IFF_BROADCAST | IFF_SIMPLEX | IFF_MULTICAST; ifp->if_init = cxgbe_init; ifp->if_ioctl = cxgbe_ioctl; ifp->if_transmit = cxgbe_transmit; ifp->if_qflush = cxgbe_qflush; ifp->if_get_counter = cxgbe_get_counter; ifp->if_capabilities = T4_CAP; #ifdef TCP_OFFLOAD if (is_offload(pi->adapter)) ifp->if_capabilities |= IFCAP_TOE; #endif ifp->if_capenable = T4_CAP_ENABLE; ifp->if_hwassist = CSUM_TCP | CSUM_UDP | CSUM_IP | CSUM_TSO | CSUM_UDP_IPV6 | CSUM_TCP_IPV6; ifp->if_hw_tsomax = 65536 - (ETHER_HDR_LEN + ETHER_VLAN_ENCAP_LEN); ifp->if_hw_tsomaxsegcount = TX_SGL_SEGS; ifp->if_hw_tsomaxsegsize = 65536; /* Initialize ifmedia for this port */ ifmedia_init(&pi->media, IFM_IMASK, cxgbe_media_change, cxgbe_media_status); build_medialist(pi, &pi->media); pi->vlan_c = EVENTHANDLER_REGISTER(vlan_config, cxgbe_vlan_config, ifp, EVENTHANDLER_PRI_ANY); ether_ifattach(ifp, pi->hw_addr); n = 128; s = malloc(n, M_CXGBE, M_WAITOK); o = snprintf(s, n, "%d txq, %d rxq (NIC)", pi->ntxq, pi->nrxq); MPASS(n > o); #ifdef TCP_OFFLOAD if (is_offload(pi->adapter)) { o += snprintf(s + o, n - o, "; %d txq, %d rxq (TOE)", pi->nofldtxq, pi->nofldrxq); MPASS(n > o); } #endif #ifdef DEV_NETMAP o += snprintf(s + o, n - o, "; %d txq, %d rxq (netmap)", pi->nnmtxq, pi->nnmrxq); MPASS(n > o); #endif device_printf(dev, "%s\n", s); free(s, M_CXGBE); #ifdef DEV_NETMAP /* nm_media handled here to keep implementation private to this file */ ifmedia_init(&pi->nm_media, IFM_IMASK, cxgbe_media_change, cxgbe_media_status); build_medialist(pi, &pi->nm_media); create_netmap_ifnet(pi); /* logs errors it something fails */ #endif cxgbe_sysctls(pi); return (0); } static int cxgbe_detach(device_t dev) { struct port_info *pi = device_get_softc(dev); struct adapter *sc = pi->adapter; struct ifnet *ifp = pi->ifp; /* Tell if_ioctl and if_init that the port is going away */ ADAPTER_LOCK(sc); SET_DOOMED(pi); wakeup(&sc->flags); while (IS_BUSY(sc)) mtx_sleep(&sc->flags, &sc->sc_lock, 0, "t4detach", 0); SET_BUSY(sc); #ifdef INVARIANTS sc->last_op = "t4detach"; sc->last_op_thr = curthread; #endif ADAPTER_UNLOCK(sc); if (pi->flags & HAS_TRACEQ) { sc->traceq = -1; /* cloner should not create ifnet */ t4_tracer_port_detach(sc); } if (pi->vlan_c) EVENTHANDLER_DEREGISTER(vlan_config, pi->vlan_c); PORT_LOCK(pi); ifp->if_drv_flags &= ~IFF_DRV_RUNNING; callout_stop(&pi->tick); PORT_UNLOCK(pi); callout_drain(&pi->tick); /* Let detach proceed even if these fail. */ cxgbe_uninit_synchronized(pi); port_full_uninit(pi); ifmedia_removeall(&pi->media); ether_ifdetach(pi->ifp); if_free(pi->ifp); #ifdef DEV_NETMAP /* XXXNM: equivalent of cxgbe_uninit_synchronized to ifdown nm_ifp */ destroy_netmap_ifnet(pi); #endif ADAPTER_LOCK(sc); CLR_BUSY(sc); wakeup(&sc->flags); ADAPTER_UNLOCK(sc); return (0); } static void cxgbe_init(void *arg) { struct port_info *pi = arg; struct adapter *sc = pi->adapter; if (begin_synchronized_op(sc, pi, SLEEP_OK | INTR_OK, "t4init") != 0) return; cxgbe_init_synchronized(pi); end_synchronized_op(sc, 0); } static int cxgbe_ioctl(struct ifnet *ifp, unsigned long cmd, caddr_t data) { int rc = 0, mtu, flags, can_sleep; struct port_info *pi = ifp->if_softc; struct adapter *sc = pi->adapter; struct ifreq *ifr = (struct ifreq *)data; uint32_t mask; switch (cmd) { case SIOCSIFMTU: mtu = ifr->ifr_mtu; if ((mtu < ETHERMIN) || (mtu > ETHERMTU_JUMBO)) return (EINVAL); rc = begin_synchronized_op(sc, pi, SLEEP_OK | INTR_OK, "t4mtu"); if (rc) return (rc); ifp->if_mtu = mtu; if (pi->flags & PORT_INIT_DONE) { t4_update_fl_bufsize(ifp); if (ifp->if_drv_flags & IFF_DRV_RUNNING) rc = update_mac_settings(ifp, XGMAC_MTU); } end_synchronized_op(sc, 0); break; case SIOCSIFFLAGS: can_sleep = 0; redo_sifflags: rc = begin_synchronized_op(sc, pi, can_sleep ? (SLEEP_OK | INTR_OK) : HOLD_LOCK, "t4flg"); if (rc) return (rc); if (ifp->if_flags & IFF_UP) { if (ifp->if_drv_flags & IFF_DRV_RUNNING) { flags = pi->if_flags; if ((ifp->if_flags ^ flags) & (IFF_PROMISC | IFF_ALLMULTI)) { if (can_sleep == 1) { end_synchronized_op(sc, 0); can_sleep = 0; goto redo_sifflags; } rc = update_mac_settings(ifp, XGMAC_PROMISC | XGMAC_ALLMULTI); } } else { if (can_sleep == 0) { end_synchronized_op(sc, LOCK_HELD); can_sleep = 1; goto redo_sifflags; } rc = cxgbe_init_synchronized(pi); } pi->if_flags = ifp->if_flags; } else if (ifp->if_drv_flags & IFF_DRV_RUNNING) { if (can_sleep == 0) { end_synchronized_op(sc, LOCK_HELD); can_sleep = 1; goto redo_sifflags; } rc = cxgbe_uninit_synchronized(pi); } end_synchronized_op(sc, can_sleep ? 0 : LOCK_HELD); break; case SIOCADDMULTI: case SIOCDELMULTI: /* these two are called with a mutex held :-( */ rc = begin_synchronized_op(sc, pi, HOLD_LOCK, "t4multi"); if (rc) return (rc); if (ifp->if_drv_flags & IFF_DRV_RUNNING) rc = update_mac_settings(ifp, XGMAC_MCADDRS); end_synchronized_op(sc, LOCK_HELD); break; case SIOCSIFCAP: rc = begin_synchronized_op(sc, pi, SLEEP_OK | INTR_OK, "t4cap"); if (rc) return (rc); mask = ifr->ifr_reqcap ^ ifp->if_capenable; if (mask & IFCAP_TXCSUM) { ifp->if_capenable ^= IFCAP_TXCSUM; ifp->if_hwassist ^= (CSUM_TCP | CSUM_UDP | CSUM_IP); if (IFCAP_TSO4 & ifp->if_capenable && !(IFCAP_TXCSUM & ifp->if_capenable)) { ifp->if_capenable &= ~IFCAP_TSO4; if_printf(ifp, "tso4 disabled due to -txcsum.\n"); } } if (mask & IFCAP_TXCSUM_IPV6) { ifp->if_capenable ^= IFCAP_TXCSUM_IPV6; ifp->if_hwassist ^= (CSUM_UDP_IPV6 | CSUM_TCP_IPV6); if (IFCAP_TSO6 & ifp->if_capenable && !(IFCAP_TXCSUM_IPV6 & ifp->if_capenable)) { ifp->if_capenable &= ~IFCAP_TSO6; if_printf(ifp, "tso6 disabled due to -txcsum6.\n"); } } if (mask & IFCAP_RXCSUM) ifp->if_capenable ^= IFCAP_RXCSUM; if (mask & IFCAP_RXCSUM_IPV6) ifp->if_capenable ^= IFCAP_RXCSUM_IPV6; /* * Note that we leave CSUM_TSO alone (it is always set). The * kernel takes both IFCAP_TSOx and CSUM_TSO into account before * sending a TSO request our way, so it's sufficient to toggle * IFCAP_TSOx only. */ if (mask & IFCAP_TSO4) { if (!(IFCAP_TSO4 & ifp->if_capenable) && !(IFCAP_TXCSUM & ifp->if_capenable)) { if_printf(ifp, "enable txcsum first.\n"); rc = EAGAIN; goto fail; } ifp->if_capenable ^= IFCAP_TSO4; } if (mask & IFCAP_TSO6) { if (!(IFCAP_TSO6 & ifp->if_capenable) && !(IFCAP_TXCSUM_IPV6 & ifp->if_capenable)) { if_printf(ifp, "enable txcsum6 first.\n"); rc = EAGAIN; goto fail; } ifp->if_capenable ^= IFCAP_TSO6; } if (mask & IFCAP_LRO) { #if defined(INET) || defined(INET6) int i; struct sge_rxq *rxq; ifp->if_capenable ^= IFCAP_LRO; for_each_rxq(pi, i, rxq) { if (ifp->if_capenable & IFCAP_LRO) rxq->iq.flags |= IQ_LRO_ENABLED; else rxq->iq.flags &= ~IQ_LRO_ENABLED; } #endif } #ifdef TCP_OFFLOAD if (mask & IFCAP_TOE) { int enable = (ifp->if_capenable ^ mask) & IFCAP_TOE; rc = toe_capability(pi, enable); if (rc != 0) goto fail; ifp->if_capenable ^= mask; } #endif if (mask & IFCAP_VLAN_HWTAGGING) { ifp->if_capenable ^= IFCAP_VLAN_HWTAGGING; if (ifp->if_drv_flags & IFF_DRV_RUNNING) rc = update_mac_settings(ifp, XGMAC_VLANEX); } if (mask & IFCAP_VLAN_MTU) { ifp->if_capenable ^= IFCAP_VLAN_MTU; /* Need to find out how to disable auto-mtu-inflation */ } if (mask & IFCAP_VLAN_HWTSO) ifp->if_capenable ^= IFCAP_VLAN_HWTSO; if (mask & IFCAP_VLAN_HWCSUM) ifp->if_capenable ^= IFCAP_VLAN_HWCSUM; #ifdef VLAN_CAPABILITIES VLAN_CAPABILITIES(ifp); #endif fail: end_synchronized_op(sc, 0); break; case SIOCSIFMEDIA: case SIOCGIFMEDIA: ifmedia_ioctl(ifp, ifr, &pi->media, cmd); break; case SIOCGI2C: { struct ifi2creq i2c; rc = copyin(ifr->ifr_data, &i2c, sizeof(i2c)); if (rc != 0) break; if (i2c.dev_addr != 0xA0 && i2c.dev_addr != 0xA2) { rc = EPERM; break; } if (i2c.len > sizeof(i2c.data)) { rc = EINVAL; break; } rc = begin_synchronized_op(sc, pi, SLEEP_OK | INTR_OK, "t4i2c"); if (rc) return (rc); rc = -t4_i2c_rd(sc, sc->mbox, pi->port_id, i2c.dev_addr, i2c.offset, i2c.len, &i2c.data[0]); end_synchronized_op(sc, 0); if (rc == 0) rc = copyout(&i2c, ifr->ifr_data, sizeof(i2c)); break; } default: rc = ether_ioctl(ifp, cmd, data); } return (rc); } static int cxgbe_transmit(struct ifnet *ifp, struct mbuf *m) { struct port_info *pi = ifp->if_softc; struct adapter *sc = pi->adapter; struct sge_txq *txq; void *items[1]; int rc; M_ASSERTPKTHDR(m); MPASS(m->m_nextpkt == NULL); /* not quite ready for this yet */ if (__predict_false(pi->link_cfg.link_ok == 0)) { m_freem(m); return (ENETDOWN); } rc = parse_pkt(&m); if (__predict_false(rc != 0)) { MPASS(m == NULL); /* was freed already */ atomic_add_int(&pi->tx_parse_error, 1); /* rare, atomic is ok */ return (rc); } /* Select a txq. */ txq = &sc->sge.txq[pi->first_txq]; if (M_HASHTYPE_GET(m) != M_HASHTYPE_NONE) txq += ((m->m_pkthdr.flowid % (pi->ntxq - pi->rsrv_noflowq)) + pi->rsrv_noflowq); items[0] = m; rc = mp_ring_enqueue(txq->r, items, 1, 4096); if (__predict_false(rc != 0)) m_freem(m); return (rc); } static void cxgbe_qflush(struct ifnet *ifp) { struct port_info *pi = ifp->if_softc; struct sge_txq *txq; int i; /* queues do not exist if !PORT_INIT_DONE. */ if (pi->flags & PORT_INIT_DONE) { for_each_txq(pi, i, txq) { TXQ_LOCK(txq); txq->eq.flags &= ~EQ_ENABLED; TXQ_UNLOCK(txq); while (!mp_ring_is_idle(txq->r)) { mp_ring_check_drainage(txq->r, 0); pause("qflush", 1); } } } if_qflush(ifp); } static uint64_t cxgbe_get_counter(struct ifnet *ifp, ift_counter c) { struct port_info *pi = ifp->if_softc; struct adapter *sc = pi->adapter; struct port_stats *s = &pi->stats; cxgbe_refresh_stats(sc, pi); switch (c) { case IFCOUNTER_IPACKETS: return (s->rx_frames - s->rx_pause); case IFCOUNTER_IERRORS: return (s->rx_jabber + s->rx_runt + s->rx_too_long + s->rx_fcs_err + s->rx_len_err); case IFCOUNTER_OPACKETS: return (s->tx_frames - s->tx_pause); case IFCOUNTER_OERRORS: return (s->tx_error_frames); case IFCOUNTER_IBYTES: return (s->rx_octets - s->rx_pause * 64); case IFCOUNTER_OBYTES: return (s->tx_octets - s->tx_pause * 64); case IFCOUNTER_IMCASTS: return (s->rx_mcast_frames - s->rx_pause); case IFCOUNTER_OMCASTS: return (s->tx_mcast_frames - s->tx_pause); case IFCOUNTER_IQDROPS: return (s->rx_ovflow0 + s->rx_ovflow1 + s->rx_ovflow2 + s->rx_ovflow3 + s->rx_trunc0 + s->rx_trunc1 + s->rx_trunc2 + s->rx_trunc3 + pi->tnl_cong_drops); case IFCOUNTER_OQDROPS: { uint64_t drops; drops = s->tx_drop; if (pi->flags & PORT_INIT_DONE) { int i; struct sge_txq *txq; for_each_txq(pi, i, txq) drops += counter_u64_fetch(txq->r->drops); } return (drops); } default: return (if_get_counter_default(ifp, c)); } } static int cxgbe_media_change(struct ifnet *ifp) { struct port_info *pi = ifp->if_softc; device_printf(pi->dev, "%s unimplemented.\n", __func__); return (EOPNOTSUPP); } static void cxgbe_media_status(struct ifnet *ifp, struct ifmediareq *ifmr) { struct port_info *pi = ifp->if_softc; struct ifmedia *media = NULL; struct ifmedia_entry *cur; int speed = pi->link_cfg.speed; if (ifp == pi->ifp) media = &pi->media; #ifdef DEV_NETMAP else if (ifp == pi->nm_ifp) media = &pi->nm_media; #endif MPASS(media != NULL); cur = media->ifm_cur; ifmr->ifm_status = IFM_AVALID; if (!pi->link_cfg.link_ok) return; ifmr->ifm_status |= IFM_ACTIVE; /* active and current will differ iff current media is autoselect. */ if (IFM_SUBTYPE(cur->ifm_media) != IFM_AUTO) return; ifmr->ifm_active = IFM_ETHER | IFM_FDX; if (speed == SPEED_10000) ifmr->ifm_active |= IFM_10G_T; else if (speed == SPEED_1000) ifmr->ifm_active |= IFM_1000_T; else if (speed == SPEED_100) ifmr->ifm_active |= IFM_100_TX; else if (speed == SPEED_10) ifmr->ifm_active |= IFM_10_T; else KASSERT(0, ("%s: link up but speed unknown (%u)", __func__, speed)); } void t4_fatal_err(struct adapter *sc) { t4_set_reg_field(sc, A_SGE_CONTROL, F_GLOBALENABLE, 0); t4_intr_disable(sc); log(LOG_EMERG, "%s: encountered fatal error, adapter stopped.\n", device_get_nameunit(sc->dev)); } static int map_bars_0_and_4(struct adapter *sc) { sc->regs_rid = PCIR_BAR(0); sc->regs_res = bus_alloc_resource_any(sc->dev, SYS_RES_MEMORY, &sc->regs_rid, RF_ACTIVE); if (sc->regs_res == NULL) { device_printf(sc->dev, "cannot map registers.\n"); return (ENXIO); } sc->bt = rman_get_bustag(sc->regs_res); sc->bh = rman_get_bushandle(sc->regs_res); sc->mmio_len = rman_get_size(sc->regs_res); setbit(&sc->doorbells, DOORBELL_KDB); sc->msix_rid = PCIR_BAR(4); sc->msix_res = bus_alloc_resource_any(sc->dev, SYS_RES_MEMORY, &sc->msix_rid, RF_ACTIVE); if (sc->msix_res == NULL) { device_printf(sc->dev, "cannot map MSI-X BAR.\n"); return (ENXIO); } return (0); } static int map_bar_2(struct adapter *sc) { /* * T4: only iWARP driver uses the userspace doorbells. There is no need * to map it if RDMA is disabled. */ if (is_t4(sc) && sc->rdmacaps == 0) return (0); sc->udbs_rid = PCIR_BAR(2); sc->udbs_res = bus_alloc_resource_any(sc->dev, SYS_RES_MEMORY, &sc->udbs_rid, RF_ACTIVE); if (sc->udbs_res == NULL) { device_printf(sc->dev, "cannot map doorbell BAR.\n"); return (ENXIO); } sc->udbs_base = rman_get_virtual(sc->udbs_res); if (is_t5(sc)) { setbit(&sc->doorbells, DOORBELL_UDB); #if defined(__i386__) || defined(__amd64__) if (t5_write_combine) { int rc; /* * Enable write combining on BAR2. This is the * userspace doorbell BAR and is split into 128B * (UDBS_SEG_SIZE) doorbell regions, each associated * with an egress queue. The first 64B has the doorbell * and the second 64B can be used to submit a tx work * request with an implicit doorbell. */ rc = pmap_change_attr((vm_offset_t)sc->udbs_base, rman_get_size(sc->udbs_res), PAT_WRITE_COMBINING); if (rc == 0) { clrbit(&sc->doorbells, DOORBELL_UDB); setbit(&sc->doorbells, DOORBELL_WCWR); setbit(&sc->doorbells, DOORBELL_UDBWC); } else { device_printf(sc->dev, "couldn't enable write combining: %d\n", rc); } t4_write_reg(sc, A_SGE_STAT_CFG, V_STATSOURCE_T5(7) | V_STATMODE(0)); } #endif } return (0); } static const struct memwin t4_memwin[] = { { MEMWIN0_BASE, MEMWIN0_APERTURE }, { MEMWIN1_BASE, MEMWIN1_APERTURE }, { MEMWIN2_BASE_T4, MEMWIN2_APERTURE_T4 } }; static const struct memwin t5_memwin[] = { { MEMWIN0_BASE, MEMWIN0_APERTURE }, { MEMWIN1_BASE, MEMWIN1_APERTURE }, { MEMWIN2_BASE_T5, MEMWIN2_APERTURE_T5 }, }; static void setup_memwin(struct adapter *sc) { const struct memwin *mw; int i, n; uint32_t bar0; if (is_t4(sc)) { /* * Read low 32b of bar0 indirectly via the hardware backdoor * mechanism. Works from within PCI passthrough environments * too, where rman_get_start() can return a different value. We * need to program the T4 memory window decoders with the actual * addresses that will be coming across the PCIe link. */ bar0 = t4_hw_pci_read_cfg4(sc, PCIR_BAR(0)); bar0 &= (uint32_t) PCIM_BAR_MEM_BASE; mw = &t4_memwin[0]; n = nitems(t4_memwin); } else { /* T5 uses the relative offset inside the PCIe BAR */ bar0 = 0; mw = &t5_memwin[0]; n = nitems(t5_memwin); } for (i = 0; i < n; i++, mw++) { t4_write_reg(sc, PCIE_MEM_ACCESS_REG(A_PCIE_MEM_ACCESS_BASE_WIN, i), (mw->base + bar0) | V_BIR(0) | V_WINDOW(ilog2(mw->aperture) - 10)); } /* flush */ t4_read_reg(sc, PCIE_MEM_ACCESS_REG(A_PCIE_MEM_ACCESS_BASE_WIN, 2)); } /* * Verify that the memory range specified by the addr/len pair is valid and lies * entirely within a single region (EDCx or MCx). */ static int validate_mem_range(struct adapter *sc, uint32_t addr, int len) { uint32_t em, addr_len, maddr, mlen; /* Memory can only be accessed in naturally aligned 4 byte units */ if (addr & 3 || len & 3 || len == 0) return (EINVAL); /* Enabled memories */ em = t4_read_reg(sc, A_MA_TARGET_MEM_ENABLE); if (em & F_EDRAM0_ENABLE) { addr_len = t4_read_reg(sc, A_MA_EDRAM0_BAR); maddr = G_EDRAM0_BASE(addr_len) << 20; mlen = G_EDRAM0_SIZE(addr_len) << 20; if (mlen > 0 && addr >= maddr && addr < maddr + mlen && addr + len <= maddr + mlen) return (0); } if (em & F_EDRAM1_ENABLE) { addr_len = t4_read_reg(sc, A_MA_EDRAM1_BAR); maddr = G_EDRAM1_BASE(addr_len) << 20; mlen = G_EDRAM1_SIZE(addr_len) << 20; if (mlen > 0 && addr >= maddr && addr < maddr + mlen && addr + len <= maddr + mlen) return (0); } if (em & F_EXT_MEM_ENABLE) { addr_len = t4_read_reg(sc, A_MA_EXT_MEMORY_BAR); maddr = G_EXT_MEM_BASE(addr_len) << 20; mlen = G_EXT_MEM_SIZE(addr_len) << 20; if (mlen > 0 && addr >= maddr && addr < maddr + mlen && addr + len <= maddr + mlen) return (0); } if (!is_t4(sc) && em & F_EXT_MEM1_ENABLE) { addr_len = t4_read_reg(sc, A_MA_EXT_MEMORY1_BAR); maddr = G_EXT_MEM1_BASE(addr_len) << 20; mlen = G_EXT_MEM1_SIZE(addr_len) << 20; if (mlen > 0 && addr >= maddr && addr < maddr + mlen && addr + len <= maddr + mlen) return (0); } return (EFAULT); } static int fwmtype_to_hwmtype(int mtype) { switch (mtype) { case FW_MEMTYPE_EDC0: return (MEM_EDC0); case FW_MEMTYPE_EDC1: return (MEM_EDC1); case FW_MEMTYPE_EXTMEM: return (MEM_MC0); case FW_MEMTYPE_EXTMEM1: return (MEM_MC1); default: panic("%s: cannot translate fw mtype %d.", __func__, mtype); } } /* * Verify that the memory range specified by the memtype/offset/len pair is * valid and lies entirely within the memtype specified. The global address of * the start of the range is returned in addr. */ static int validate_mt_off_len(struct adapter *sc, int mtype, uint32_t off, int len, uint32_t *addr) { uint32_t em, addr_len, maddr, mlen; /* Memory can only be accessed in naturally aligned 4 byte units */ if (off & 3 || len & 3 || len == 0) return (EINVAL); em = t4_read_reg(sc, A_MA_TARGET_MEM_ENABLE); switch (fwmtype_to_hwmtype(mtype)) { case MEM_EDC0: if (!(em & F_EDRAM0_ENABLE)) return (EINVAL); addr_len = t4_read_reg(sc, A_MA_EDRAM0_BAR); maddr = G_EDRAM0_BASE(addr_len) << 20; mlen = G_EDRAM0_SIZE(addr_len) << 20; break; case MEM_EDC1: if (!(em & F_EDRAM1_ENABLE)) return (EINVAL); addr_len = t4_read_reg(sc, A_MA_EDRAM1_BAR); maddr = G_EDRAM1_BASE(addr_len) << 20; mlen = G_EDRAM1_SIZE(addr_len) << 20; break; case MEM_MC: if (!(em & F_EXT_MEM_ENABLE)) return (EINVAL); addr_len = t4_read_reg(sc, A_MA_EXT_MEMORY_BAR); maddr = G_EXT_MEM_BASE(addr_len) << 20; mlen = G_EXT_MEM_SIZE(addr_len) << 20; break; case MEM_MC1: if (is_t4(sc) || !(em & F_EXT_MEM1_ENABLE)) return (EINVAL); addr_len = t4_read_reg(sc, A_MA_EXT_MEMORY1_BAR); maddr = G_EXT_MEM1_BASE(addr_len) << 20; mlen = G_EXT_MEM1_SIZE(addr_len) << 20; break; default: return (EINVAL); } if (mlen > 0 && off < mlen && off + len <= mlen) { *addr = maddr + off; /* global address */ return (0); } return (EFAULT); } static void memwin_info(struct adapter *sc, int win, uint32_t *base, uint32_t *aperture) { const struct memwin *mw; if (is_t4(sc)) { KASSERT(win >= 0 && win < nitems(t4_memwin), ("%s: incorrect memwin# (%d)", __func__, win)); mw = &t4_memwin[win]; } else { KASSERT(win >= 0 && win < nitems(t5_memwin), ("%s: incorrect memwin# (%d)", __func__, win)); mw = &t5_memwin[win]; } if (base != NULL) *base = mw->base; if (aperture != NULL) *aperture = mw->aperture; } /* * Positions the memory window such that it can be used to access the specified * address in the chip's address space. The return value is the offset of addr * from the start of the window. */ static uint32_t position_memwin(struct adapter *sc, int n, uint32_t addr) { uint32_t start, pf; uint32_t reg; KASSERT(n >= 0 && n <= 3, ("%s: invalid window %d.", __func__, n)); KASSERT((addr & 3) == 0, ("%s: addr (0x%x) is not at a 4B boundary.", __func__, addr)); if (is_t4(sc)) { pf = 0; start = addr & ~0xf; /* start must be 16B aligned */ } else { pf = V_PFNUM(sc->pf); start = addr & ~0x7f; /* start must be 128B aligned */ } reg = PCIE_MEM_ACCESS_REG(A_PCIE_MEM_ACCESS_OFFSET, n); t4_write_reg(sc, reg, start | pf); t4_read_reg(sc, reg); return (addr - start); } static int cfg_itype_and_nqueues(struct adapter *sc, int n10g, int n1g, struct intrs_and_queues *iaq) { int rc, itype, navail, nrxq10g, nrxq1g, n; int nofldrxq10g = 0, nofldrxq1g = 0; int nnmrxq10g = 0, nnmrxq1g = 0; bzero(iaq, sizeof(*iaq)); iaq->ntxq10g = t4_ntxq10g; iaq->ntxq1g = t4_ntxq1g; iaq->nrxq10g = nrxq10g = t4_nrxq10g; iaq->nrxq1g = nrxq1g = t4_nrxq1g; iaq->rsrv_noflowq = t4_rsrv_noflowq; #ifdef TCP_OFFLOAD if (is_offload(sc)) { iaq->nofldtxq10g = t4_nofldtxq10g; iaq->nofldtxq1g = t4_nofldtxq1g; iaq->nofldrxq10g = nofldrxq10g = t4_nofldrxq10g; iaq->nofldrxq1g = nofldrxq1g = t4_nofldrxq1g; } #endif #ifdef DEV_NETMAP iaq->nnmtxq10g = t4_nnmtxq10g; iaq->nnmtxq1g = t4_nnmtxq1g; iaq->nnmrxq10g = nnmrxq10g = t4_nnmrxq10g; iaq->nnmrxq1g = nnmrxq1g = t4_nnmrxq1g; #endif for (itype = INTR_MSIX; itype; itype >>= 1) { if ((itype & t4_intr_types) == 0) continue; /* not allowed */ if (itype == INTR_MSIX) navail = pci_msix_count(sc->dev); else if (itype == INTR_MSI) navail = pci_msi_count(sc->dev); else navail = 1; restart: if (navail == 0) continue; iaq->intr_type = itype; iaq->intr_flags_10g = 0; iaq->intr_flags_1g = 0; /* * Best option: an interrupt vector for errors, one for the * firmware event queue, and one for every rxq (NIC, TOE, and * netmap). */ iaq->nirq = T4_EXTRA_INTR; iaq->nirq += n10g * (nrxq10g + nofldrxq10g + nnmrxq10g); iaq->nirq += n1g * (nrxq1g + nofldrxq1g + nnmrxq1g); if (iaq->nirq <= navail && (itype != INTR_MSI || powerof2(iaq->nirq))) { iaq->intr_flags_10g = INTR_ALL; iaq->intr_flags_1g = INTR_ALL; goto allocate; } /* * Second best option: a vector for errors, one for the firmware * event queue, and vectors for either all the NIC rx queues or * all the TOE rx queues. The queues that don't get vectors * will forward their interrupts to those that do. * * Note: netmap rx queues cannot be created early and so they * can't be setup to receive forwarded interrupts for others. */ iaq->nirq = T4_EXTRA_INTR; if (nrxq10g >= nofldrxq10g) { iaq->intr_flags_10g = INTR_RXQ; iaq->nirq += n10g * nrxq10g; #ifdef DEV_NETMAP iaq->nnmrxq10g = min(nnmrxq10g, nrxq10g); #endif } else { iaq->intr_flags_10g = INTR_OFLD_RXQ; iaq->nirq += n10g * nofldrxq10g; #ifdef DEV_NETMAP iaq->nnmrxq10g = min(nnmrxq10g, nofldrxq10g); #endif } if (nrxq1g >= nofldrxq1g) { iaq->intr_flags_1g = INTR_RXQ; iaq->nirq += n1g * nrxq1g; #ifdef DEV_NETMAP iaq->nnmrxq1g = min(nnmrxq1g, nrxq1g); #endif } else { iaq->intr_flags_1g = INTR_OFLD_RXQ; iaq->nirq += n1g * nofldrxq1g; #ifdef DEV_NETMAP iaq->nnmrxq1g = min(nnmrxq1g, nofldrxq1g); #endif } if (iaq->nirq <= navail && (itype != INTR_MSI || powerof2(iaq->nirq))) goto allocate; /* * Next best option: an interrupt vector for errors, one for the * firmware event queue, and at least one per port. At this * point we know we'll have to downsize nrxq and/or nofldrxq * and/or nnmrxq to fit what's available to us. */ iaq->nirq = T4_EXTRA_INTR; iaq->nirq += n10g + n1g; if (iaq->nirq <= navail) { int leftover = navail - iaq->nirq; if (n10g > 0) { int target = max(nrxq10g, nofldrxq10g); iaq->intr_flags_10g = nrxq10g >= nofldrxq10g ? INTR_RXQ : INTR_OFLD_RXQ; n = 1; while (n < target && leftover >= n10g) { leftover -= n10g; iaq->nirq += n10g; n++; } iaq->nrxq10g = min(n, nrxq10g); #ifdef TCP_OFFLOAD iaq->nofldrxq10g = min(n, nofldrxq10g); #endif #ifdef DEV_NETMAP iaq->nnmrxq10g = min(n, nnmrxq10g); #endif } if (n1g > 0) { int target = max(nrxq1g, nofldrxq1g); iaq->intr_flags_1g = nrxq1g >= nofldrxq1g ? INTR_RXQ : INTR_OFLD_RXQ; n = 1; while (n < target && leftover >= n1g) { leftover -= n1g; iaq->nirq += n1g; n++; } iaq->nrxq1g = min(n, nrxq1g); #ifdef TCP_OFFLOAD iaq->nofldrxq1g = min(n, nofldrxq1g); #endif #ifdef DEV_NETMAP iaq->nnmrxq1g = min(n, nnmrxq1g); #endif } if (itype != INTR_MSI || powerof2(iaq->nirq)) goto allocate; } /* * Least desirable option: one interrupt vector for everything. */ iaq->nirq = iaq->nrxq10g = iaq->nrxq1g = 1; iaq->intr_flags_10g = iaq->intr_flags_1g = 0; #ifdef TCP_OFFLOAD if (is_offload(sc)) iaq->nofldrxq10g = iaq->nofldrxq1g = 1; #endif #ifdef DEV_NETMAP iaq->nnmrxq10g = iaq->nnmrxq1g = 1; #endif allocate: navail = iaq->nirq; rc = 0; if (itype == INTR_MSIX) rc = pci_alloc_msix(sc->dev, &navail); else if (itype == INTR_MSI) rc = pci_alloc_msi(sc->dev, &navail); if (rc == 0) { if (navail == iaq->nirq) return (0); /* * Didn't get the number requested. Use whatever number * the kernel is willing to allocate (it's in navail). */ device_printf(sc->dev, "fewer vectors than requested, " "type=%d, req=%d, rcvd=%d; will downshift req.\n", itype, iaq->nirq, navail); pci_release_msi(sc->dev); goto restart; } device_printf(sc->dev, "failed to allocate vectors:%d, type=%d, req=%d, rcvd=%d\n", itype, rc, iaq->nirq, navail); } device_printf(sc->dev, "failed to find a usable interrupt type. " "allowed=%d, msi-x=%d, msi=%d, intx=1", t4_intr_types, pci_msix_count(sc->dev), pci_msi_count(sc->dev)); return (ENXIO); } #define FW_VERSION(chip) ( \ V_FW_HDR_FW_VER_MAJOR(chip##FW_VERSION_MAJOR) | \ V_FW_HDR_FW_VER_MINOR(chip##FW_VERSION_MINOR) | \ V_FW_HDR_FW_VER_MICRO(chip##FW_VERSION_MICRO) | \ V_FW_HDR_FW_VER_BUILD(chip##FW_VERSION_BUILD)) #define FW_INTFVER(chip, intf) (chip##FW_HDR_INTFVER_##intf) struct fw_info { uint8_t chip; char *kld_name; char *fw_mod_name; struct fw_hdr fw_hdr; /* XXX: waste of space, need a sparse struct */ } fw_info[] = { { .chip = CHELSIO_T4, .kld_name = "t4fw_cfg", .fw_mod_name = "t4fw", .fw_hdr = { .chip = FW_HDR_CHIP_T4, .fw_ver = htobe32_const(FW_VERSION(T4)), .intfver_nic = FW_INTFVER(T4, NIC), .intfver_vnic = FW_INTFVER(T4, VNIC), .intfver_ofld = FW_INTFVER(T4, OFLD), .intfver_ri = FW_INTFVER(T4, RI), .intfver_iscsipdu = FW_INTFVER(T4, ISCSIPDU), .intfver_iscsi = FW_INTFVER(T4, ISCSI), .intfver_fcoepdu = FW_INTFVER(T4, FCOEPDU), .intfver_fcoe = FW_INTFVER(T4, FCOE), }, }, { .chip = CHELSIO_T5, .kld_name = "t5fw_cfg", .fw_mod_name = "t5fw", .fw_hdr = { .chip = FW_HDR_CHIP_T5, .fw_ver = htobe32_const(FW_VERSION(T5)), .intfver_nic = FW_INTFVER(T5, NIC), .intfver_vnic = FW_INTFVER(T5, VNIC), .intfver_ofld = FW_INTFVER(T5, OFLD), .intfver_ri = FW_INTFVER(T5, RI), .intfver_iscsipdu = FW_INTFVER(T5, ISCSIPDU), .intfver_iscsi = FW_INTFVER(T5, ISCSI), .intfver_fcoepdu = FW_INTFVER(T5, FCOEPDU), .intfver_fcoe = FW_INTFVER(T5, FCOE), }, } }; static struct fw_info * find_fw_info(int chip) { int i; for (i = 0; i < nitems(fw_info); i++) { if (fw_info[i].chip == chip) return (&fw_info[i]); } return (NULL); } /* * Is the given firmware API compatible with the one the driver was compiled * with? */ static int fw_compatible(const struct fw_hdr *hdr1, const struct fw_hdr *hdr2) { /* short circuit if it's the exact same firmware version */ if (hdr1->chip == hdr2->chip && hdr1->fw_ver == hdr2->fw_ver) return (1); /* * XXX: Is this too conservative? Perhaps I should limit this to the * features that are supported in the driver. */ #define SAME_INTF(x) (hdr1->intfver_##x == hdr2->intfver_##x) if (hdr1->chip == hdr2->chip && SAME_INTF(nic) && SAME_INTF(vnic) && SAME_INTF(ofld) && SAME_INTF(ri) && SAME_INTF(iscsipdu) && SAME_INTF(iscsi) && SAME_INTF(fcoepdu) && SAME_INTF(fcoe)) return (1); #undef SAME_INTF return (0); } /* * The firmware in the KLD is usable, but should it be installed? This routine * explains itself in detail if it indicates the KLD firmware should be * installed. */ static int should_install_kld_fw(struct adapter *sc, int card_fw_usable, int k, int c) { const char *reason; if (!card_fw_usable) { reason = "incompatible or unusable"; goto install; } if (k > c) { reason = "older than the version bundled with this driver"; goto install; } if (t4_fw_install == 2 && k != c) { reason = "different than the version bundled with this driver"; goto install; } return (0); install: if (t4_fw_install == 0) { device_printf(sc->dev, "firmware on card (%u.%u.%u.%u) is %s, " "but the driver is prohibited from installing a different " "firmware on the card.\n", G_FW_HDR_FW_VER_MAJOR(c), G_FW_HDR_FW_VER_MINOR(c), G_FW_HDR_FW_VER_MICRO(c), G_FW_HDR_FW_VER_BUILD(c), reason); return (0); } device_printf(sc->dev, "firmware on card (%u.%u.%u.%u) is %s, " "installing firmware %u.%u.%u.%u on card.\n", G_FW_HDR_FW_VER_MAJOR(c), G_FW_HDR_FW_VER_MINOR(c), G_FW_HDR_FW_VER_MICRO(c), G_FW_HDR_FW_VER_BUILD(c), reason, G_FW_HDR_FW_VER_MAJOR(k), G_FW_HDR_FW_VER_MINOR(k), G_FW_HDR_FW_VER_MICRO(k), G_FW_HDR_FW_VER_BUILD(k)); return (1); } /* * Establish contact with the firmware and determine if we are the master driver * or not, and whether we are responsible for chip initialization. */ static int prep_firmware(struct adapter *sc) { const struct firmware *fw = NULL, *default_cfg; int rc, pf, card_fw_usable, kld_fw_usable, need_fw_reset = 1; enum dev_state state; struct fw_info *fw_info; struct fw_hdr *card_fw; /* fw on the card */ const struct fw_hdr *kld_fw; /* fw in the KLD */ const struct fw_hdr *drv_fw; /* fw header the driver was compiled against */ /* Contact firmware. */ rc = t4_fw_hello(sc, sc->mbox, sc->mbox, MASTER_MAY, &state); if (rc < 0 || state == DEV_STATE_ERR) { rc = -rc; device_printf(sc->dev, "failed to connect to the firmware: %d, %d.\n", rc, state); return (rc); } pf = rc; if (pf == sc->mbox) sc->flags |= MASTER_PF; else if (state == DEV_STATE_UNINIT) { /* * We didn't get to be the master so we definitely won't be * configuring the chip. It's a bug if someone else hasn't * configured it already. */ device_printf(sc->dev, "couldn't be master(%d), " "device not already initialized either(%d).\n", rc, state); return (EDOOFUS); } /* This is the firmware whose headers the driver was compiled against */ fw_info = find_fw_info(chip_id(sc)); if (fw_info == NULL) { device_printf(sc->dev, "unable to look up firmware information for chip %d.\n", chip_id(sc)); return (EINVAL); } drv_fw = &fw_info->fw_hdr; /* * The firmware KLD contains many modules. The KLD name is also the * name of the module that contains the default config file. */ default_cfg = firmware_get(fw_info->kld_name); /* Read the header of the firmware on the card */ card_fw = malloc(sizeof(*card_fw), M_CXGBE, M_ZERO | M_WAITOK); rc = -t4_read_flash(sc, FLASH_FW_START, sizeof (*card_fw) / sizeof (uint32_t), (uint32_t *)card_fw, 1); if (rc == 0) card_fw_usable = fw_compatible(drv_fw, (const void*)card_fw); else { device_printf(sc->dev, "Unable to read card's firmware header: %d\n", rc); card_fw_usable = 0; } /* This is the firmware in the KLD */ fw = firmware_get(fw_info->fw_mod_name); if (fw != NULL) { kld_fw = (const void *)fw->data; kld_fw_usable = fw_compatible(drv_fw, kld_fw); } else { kld_fw = NULL; kld_fw_usable = 0; } if (card_fw_usable && card_fw->fw_ver == drv_fw->fw_ver && (!kld_fw_usable || kld_fw->fw_ver == drv_fw->fw_ver)) { /* * Common case: the firmware on the card is an exact match and * the KLD is an exact match too, or the KLD is * absent/incompatible. Note that t4_fw_install = 2 is ignored * here -- use cxgbetool loadfw if you want to reinstall the * same firmware as the one on the card. */ } else if (kld_fw_usable && state == DEV_STATE_UNINIT && should_install_kld_fw(sc, card_fw_usable, be32toh(kld_fw->fw_ver), be32toh(card_fw->fw_ver))) { rc = -t4_fw_upgrade(sc, sc->mbox, fw->data, fw->datasize, 0); if (rc != 0) { device_printf(sc->dev, "failed to install firmware: %d\n", rc); goto done; } /* Installed successfully, update the cached header too. */ memcpy(card_fw, kld_fw, sizeof(*card_fw)); card_fw_usable = 1; need_fw_reset = 0; /* already reset as part of load_fw */ } if (!card_fw_usable) { uint32_t d, c, k; d = ntohl(drv_fw->fw_ver); c = ntohl(card_fw->fw_ver); k = kld_fw ? ntohl(kld_fw->fw_ver) : 0; device_printf(sc->dev, "Cannot find a usable firmware: " "fw_install %d, chip state %d, " "driver compiled with %d.%d.%d.%d, " "card has %d.%d.%d.%d, KLD has %d.%d.%d.%d\n", t4_fw_install, state, G_FW_HDR_FW_VER_MAJOR(d), G_FW_HDR_FW_VER_MINOR(d), G_FW_HDR_FW_VER_MICRO(d), G_FW_HDR_FW_VER_BUILD(d), G_FW_HDR_FW_VER_MAJOR(c), G_FW_HDR_FW_VER_MINOR(c), G_FW_HDR_FW_VER_MICRO(c), G_FW_HDR_FW_VER_BUILD(c), G_FW_HDR_FW_VER_MAJOR(k), G_FW_HDR_FW_VER_MINOR(k), G_FW_HDR_FW_VER_MICRO(k), G_FW_HDR_FW_VER_BUILD(k)); rc = EINVAL; goto done; } /* We're using whatever's on the card and it's known to be good. */ sc->params.fw_vers = ntohl(card_fw->fw_ver); snprintf(sc->fw_version, sizeof(sc->fw_version), "%u.%u.%u.%u", G_FW_HDR_FW_VER_MAJOR(sc->params.fw_vers), G_FW_HDR_FW_VER_MINOR(sc->params.fw_vers), G_FW_HDR_FW_VER_MICRO(sc->params.fw_vers), G_FW_HDR_FW_VER_BUILD(sc->params.fw_vers)); t4_get_tp_version(sc, &sc->params.tp_vers); /* Reset device */ if (need_fw_reset && (rc = -t4_fw_reset(sc, sc->mbox, F_PIORSTMODE | F_PIORST)) != 0) { device_printf(sc->dev, "firmware reset failed: %d.\n", rc); if (rc != ETIMEDOUT && rc != EIO) t4_fw_bye(sc, sc->mbox); goto done; } sc->flags |= FW_OK; rc = get_params__pre_init(sc); if (rc != 0) goto done; /* error message displayed already */ /* Partition adapter resources as specified in the config file. */ if (state == DEV_STATE_UNINIT) { KASSERT(sc->flags & MASTER_PF, ("%s: trying to change chip settings when not master.", __func__)); rc = partition_resources(sc, default_cfg, fw_info->kld_name); if (rc != 0) goto done; /* error message displayed already */ t4_tweak_chip_settings(sc); /* get basic stuff going */ rc = -t4_fw_initialize(sc, sc->mbox); if (rc != 0) { device_printf(sc->dev, "fw init failed: %d.\n", rc); goto done; } } else { snprintf(sc->cfg_file, sizeof(sc->cfg_file), "pf%d", pf); sc->cfcsum = 0; } done: free(card_fw, M_CXGBE); if (fw != NULL) firmware_put(fw, FIRMWARE_UNLOAD); if (default_cfg != NULL) firmware_put(default_cfg, FIRMWARE_UNLOAD); return (rc); } #define FW_PARAM_DEV(param) \ (V_FW_PARAMS_MNEM(FW_PARAMS_MNEM_DEV) | \ V_FW_PARAMS_PARAM_X(FW_PARAMS_PARAM_DEV_##param)) #define FW_PARAM_PFVF(param) \ (V_FW_PARAMS_MNEM(FW_PARAMS_MNEM_PFVF) | \ V_FW_PARAMS_PARAM_X(FW_PARAMS_PARAM_PFVF_##param)) /* * Partition chip resources for use between various PFs, VFs, etc. */ static int partition_resources(struct adapter *sc, const struct firmware *default_cfg, const char *name_prefix) { const struct firmware *cfg = NULL; int rc = 0; struct fw_caps_config_cmd caps; uint32_t mtype, moff, finicsum, cfcsum; /* * Figure out what configuration file to use. Pick the default config * file for the card if the user hasn't specified one explicitly. */ snprintf(sc->cfg_file, sizeof(sc->cfg_file), "%s", t4_cfg_file); if (strncmp(t4_cfg_file, DEFAULT_CF, sizeof(t4_cfg_file)) == 0) { /* Card specific overrides go here. */ if (pci_get_device(sc->dev) == 0x440a) snprintf(sc->cfg_file, sizeof(sc->cfg_file), UWIRE_CF); if (is_fpga(sc)) snprintf(sc->cfg_file, sizeof(sc->cfg_file), FPGA_CF); } /* * We need to load another module if the profile is anything except * "default" or "flash". */ if (strncmp(sc->cfg_file, DEFAULT_CF, sizeof(sc->cfg_file)) != 0 && strncmp(sc->cfg_file, FLASH_CF, sizeof(sc->cfg_file)) != 0) { char s[32]; snprintf(s, sizeof(s), "%s_%s", name_prefix, sc->cfg_file); cfg = firmware_get(s); if (cfg == NULL) { if (default_cfg != NULL) { device_printf(sc->dev, "unable to load module \"%s\" for " "configuration profile \"%s\", will use " "the default config file instead.\n", s, sc->cfg_file); snprintf(sc->cfg_file, sizeof(sc->cfg_file), "%s", DEFAULT_CF); } else { device_printf(sc->dev, "unable to load module \"%s\" for " "configuration profile \"%s\", will use " "the config file on the card's flash " "instead.\n", s, sc->cfg_file); snprintf(sc->cfg_file, sizeof(sc->cfg_file), "%s", FLASH_CF); } } } if (strncmp(sc->cfg_file, DEFAULT_CF, sizeof(sc->cfg_file)) == 0 && default_cfg == NULL) { device_printf(sc->dev, "default config file not available, will use the config " "file on the card's flash instead.\n"); snprintf(sc->cfg_file, sizeof(sc->cfg_file), "%s", FLASH_CF); } if (strncmp(sc->cfg_file, FLASH_CF, sizeof(sc->cfg_file)) != 0) { u_int cflen, i, n; const uint32_t *cfdata; uint32_t param, val, addr, off, mw_base, mw_aperture; KASSERT(cfg != NULL || default_cfg != NULL, ("%s: no config to upload", __func__)); /* * Ask the firmware where it wants us to upload the config file. */ param = FW_PARAM_DEV(CF); rc = -t4_query_params(sc, sc->mbox, sc->pf, 0, 1, ¶m, &val); if (rc != 0) { /* No support for config file? Shouldn't happen. */ device_printf(sc->dev, "failed to query config file location: %d.\n", rc); goto done; } mtype = G_FW_PARAMS_PARAM_Y(val); moff = G_FW_PARAMS_PARAM_Z(val) << 16; /* * XXX: sheer laziness. We deliberately added 4 bytes of * useless stuffing/comments at the end of the config file so * it's ok to simply throw away the last remaining bytes when * the config file is not an exact multiple of 4. This also * helps with the validate_mt_off_len check. */ if (cfg != NULL) { cflen = cfg->datasize & ~3; cfdata = cfg->data; } else { cflen = default_cfg->datasize & ~3; cfdata = default_cfg->data; } if (cflen > FLASH_CFG_MAX_SIZE) { device_printf(sc->dev, "config file too long (%d, max allowed is %d). " "Will try to use the config on the card, if any.\n", cflen, FLASH_CFG_MAX_SIZE); goto use_config_on_flash; } rc = validate_mt_off_len(sc, mtype, moff, cflen, &addr); if (rc != 0) { device_printf(sc->dev, "%s: addr (%d/0x%x) or len %d is not valid: %d. " "Will try to use the config on the card, if any.\n", __func__, mtype, moff, cflen, rc); goto use_config_on_flash; } memwin_info(sc, 2, &mw_base, &mw_aperture); while (cflen) { off = position_memwin(sc, 2, addr); n = min(cflen, mw_aperture - off); for (i = 0; i < n; i += 4) t4_write_reg(sc, mw_base + off + i, *cfdata++); cflen -= n; addr += n; } } else { use_config_on_flash: mtype = FW_MEMTYPE_FLASH; moff = t4_flash_cfg_addr(sc); } bzero(&caps, sizeof(caps)); caps.op_to_write = htobe32(V_FW_CMD_OP(FW_CAPS_CONFIG_CMD) | F_FW_CMD_REQUEST | F_FW_CMD_READ); caps.cfvalid_to_len16 = htobe32(F_FW_CAPS_CONFIG_CMD_CFVALID | V_FW_CAPS_CONFIG_CMD_MEMTYPE_CF(mtype) | V_FW_CAPS_CONFIG_CMD_MEMADDR64K_CF(moff >> 16) | FW_LEN16(caps)); rc = -t4_wr_mbox(sc, sc->mbox, &caps, sizeof(caps), &caps); if (rc != 0) { device_printf(sc->dev, "failed to pre-process config file: %d " "(mtype %d, moff 0x%x).\n", rc, mtype, moff); goto done; } finicsum = be32toh(caps.finicsum); cfcsum = be32toh(caps.cfcsum); if (finicsum != cfcsum) { device_printf(sc->dev, "WARNING: config file checksum mismatch: %08x %08x\n", finicsum, cfcsum); } sc->cfcsum = cfcsum; #define LIMIT_CAPS(x) do { \ caps.x &= htobe16(t4_##x##_allowed); \ } while (0) /* * Let the firmware know what features will (not) be used so it can tune * things accordingly. */ LIMIT_CAPS(linkcaps); LIMIT_CAPS(niccaps); LIMIT_CAPS(toecaps); LIMIT_CAPS(rdmacaps); LIMIT_CAPS(iscsicaps); LIMIT_CAPS(fcoecaps); #undef LIMIT_CAPS caps.op_to_write = htobe32(V_FW_CMD_OP(FW_CAPS_CONFIG_CMD) | F_FW_CMD_REQUEST | F_FW_CMD_WRITE); caps.cfvalid_to_len16 = htobe32(FW_LEN16(caps)); rc = -t4_wr_mbox(sc, sc->mbox, &caps, sizeof(caps), NULL); if (rc != 0) { device_printf(sc->dev, "failed to process config file: %d.\n", rc); } done: if (cfg != NULL) firmware_put(cfg, FIRMWARE_UNLOAD); return (rc); } /* * Retrieve parameters that are needed (or nice to have) very early. */ static int get_params__pre_init(struct adapter *sc) { int rc; uint32_t param[2], val[2]; struct fw_devlog_cmd cmd; struct devlog_params *dlog = &sc->params.devlog; param[0] = FW_PARAM_DEV(PORTVEC); param[1] = FW_PARAM_DEV(CCLK); rc = -t4_query_params(sc, sc->mbox, sc->pf, 0, 2, param, val); if (rc != 0) { device_printf(sc->dev, "failed to query parameters (pre_init): %d.\n", rc); return (rc); } sc->params.portvec = val[0]; sc->params.nports = bitcount32(val[0]); sc->params.vpd.cclk = val[1]; /* Read device log parameters. */ bzero(&cmd, sizeof(cmd)); cmd.op_to_write = htobe32(V_FW_CMD_OP(FW_DEVLOG_CMD) | F_FW_CMD_REQUEST | F_FW_CMD_READ); cmd.retval_len16 = htobe32(FW_LEN16(cmd)); rc = -t4_wr_mbox(sc, sc->mbox, &cmd, sizeof(cmd), &cmd); if (rc != 0) { device_printf(sc->dev, "failed to get devlog parameters: %d.\n", rc); bzero(dlog, sizeof (*dlog)); rc = 0; /* devlog isn't critical for device operation */ } else { val[0] = be32toh(cmd.memtype_devlog_memaddr16_devlog); dlog->memtype = G_FW_DEVLOG_CMD_MEMTYPE_DEVLOG(val[0]); dlog->start = G_FW_DEVLOG_CMD_MEMADDR16_DEVLOG(val[0]) << 4; dlog->size = be32toh(cmd.memsize_devlog); } return (rc); } /* * Retrieve various parameters that are of interest to the driver. The device * has been initialized by the firmware at this point. */ static int get_params__post_init(struct adapter *sc) { int rc; uint32_t param[7], val[7]; struct fw_caps_config_cmd caps; param[0] = FW_PARAM_PFVF(IQFLINT_START); param[1] = FW_PARAM_PFVF(EQ_START); param[2] = FW_PARAM_PFVF(FILTER_START); param[3] = FW_PARAM_PFVF(FILTER_END); param[4] = FW_PARAM_PFVF(L2T_START); param[5] = FW_PARAM_PFVF(L2T_END); rc = -t4_query_params(sc, sc->mbox, sc->pf, 0, 6, param, val); if (rc != 0) { device_printf(sc->dev, "failed to query parameters (post_init): %d.\n", rc); return (rc); } sc->sge.iq_start = val[0]; sc->sge.eq_start = val[1]; sc->tids.ftid_base = val[2]; sc->tids.nftids = val[3] - val[2] + 1; sc->params.ftid_min = val[2]; sc->params.ftid_max = val[3]; sc->vres.l2t.start = val[4]; sc->vres.l2t.size = val[5] - val[4] + 1; KASSERT(sc->vres.l2t.size <= L2T_SIZE, ("%s: L2 table size (%u) larger than expected (%u)", __func__, sc->vres.l2t.size, L2T_SIZE)); /* get capabilites */ bzero(&caps, sizeof(caps)); caps.op_to_write = htobe32(V_FW_CMD_OP(FW_CAPS_CONFIG_CMD) | F_FW_CMD_REQUEST | F_FW_CMD_READ); caps.cfvalid_to_len16 = htobe32(FW_LEN16(caps)); rc = -t4_wr_mbox(sc, sc->mbox, &caps, sizeof(caps), &caps); if (rc != 0) { device_printf(sc->dev, "failed to get card capabilities: %d.\n", rc); return (rc); } #define READ_CAPS(x) do { \ sc->x = htobe16(caps.x); \ } while (0) READ_CAPS(linkcaps); READ_CAPS(niccaps); READ_CAPS(toecaps); READ_CAPS(rdmacaps); READ_CAPS(iscsicaps); READ_CAPS(fcoecaps); if (sc->niccaps & FW_CAPS_CONFIG_NIC_ETHOFLD) { param[0] = FW_PARAM_PFVF(ETHOFLD_START); param[1] = FW_PARAM_PFVF(ETHOFLD_END); param[2] = FW_PARAM_DEV(FLOWC_BUFFIFO_SZ); rc = -t4_query_params(sc, sc->mbox, sc->pf, 0, 3, param, val); if (rc != 0) { device_printf(sc->dev, "failed to query NIC parameters: %d.\n", rc); return (rc); } sc->tids.etid_base = val[0]; sc->params.etid_min = val[0]; sc->tids.netids = val[1] - val[0] + 1; sc->params.netids = sc->tids.netids; sc->params.eo_wr_cred = val[2]; sc->params.ethoffload = 1; } if (sc->toecaps) { /* query offload-related parameters */ param[0] = FW_PARAM_DEV(NTID); param[1] = FW_PARAM_PFVF(SERVER_START); param[2] = FW_PARAM_PFVF(SERVER_END); param[3] = FW_PARAM_PFVF(TDDP_START); param[4] = FW_PARAM_PFVF(TDDP_END); param[5] = FW_PARAM_DEV(FLOWC_BUFFIFO_SZ); rc = -t4_query_params(sc, sc->mbox, sc->pf, 0, 6, param, val); if (rc != 0) { device_printf(sc->dev, "failed to query TOE parameters: %d.\n", rc); return (rc); } sc->tids.ntids = val[0]; sc->tids.natids = min(sc->tids.ntids / 2, MAX_ATIDS); sc->tids.stid_base = val[1]; sc->tids.nstids = val[2] - val[1] + 1; sc->vres.ddp.start = val[3]; sc->vres.ddp.size = val[4] - val[3] + 1; sc->params.ofldq_wr_cred = val[5]; sc->params.offload = 1; } if (sc->rdmacaps) { param[0] = FW_PARAM_PFVF(STAG_START); param[1] = FW_PARAM_PFVF(STAG_END); param[2] = FW_PARAM_PFVF(RQ_START); param[3] = FW_PARAM_PFVF(RQ_END); param[4] = FW_PARAM_PFVF(PBL_START); param[5] = FW_PARAM_PFVF(PBL_END); rc = -t4_query_params(sc, sc->mbox, sc->pf, 0, 6, param, val); if (rc != 0) { device_printf(sc->dev, "failed to query RDMA parameters(1): %d.\n", rc); return (rc); } sc->vres.stag.start = val[0]; sc->vres.stag.size = val[1] - val[0] + 1; sc->vres.rq.start = val[2]; sc->vres.rq.size = val[3] - val[2] + 1; sc->vres.pbl.start = val[4]; sc->vres.pbl.size = val[5] - val[4] + 1; param[0] = FW_PARAM_PFVF(SQRQ_START); param[1] = FW_PARAM_PFVF(SQRQ_END); param[2] = FW_PARAM_PFVF(CQ_START); param[3] = FW_PARAM_PFVF(CQ_END); param[4] = FW_PARAM_PFVF(OCQ_START); param[5] = FW_PARAM_PFVF(OCQ_END); rc = -t4_query_params(sc, sc->mbox, sc->pf, 0, 6, param, val); if (rc != 0) { device_printf(sc->dev, "failed to query RDMA parameters(2): %d.\n", rc); return (rc); } sc->vres.qp.start = val[0]; sc->vres.qp.size = val[1] - val[0] + 1; sc->vres.cq.start = val[2]; sc->vres.cq.size = val[3] - val[2] + 1; sc->vres.ocq.start = val[4]; sc->vres.ocq.size = val[5] - val[4] + 1; } if (sc->iscsicaps) { param[0] = FW_PARAM_PFVF(ISCSI_START); param[1] = FW_PARAM_PFVF(ISCSI_END); rc = -t4_query_params(sc, sc->mbox, sc->pf, 0, 2, param, val); if (rc != 0) { device_printf(sc->dev, "failed to query iSCSI parameters: %d.\n", rc); return (rc); } sc->vres.iscsi.start = val[0]; sc->vres.iscsi.size = val[1] - val[0] + 1; } /* * We've got the params we wanted to query via the firmware. Now grab * some others directly from the chip. */ rc = t4_read_chip_settings(sc); return (rc); } static int set_params__post_init(struct adapter *sc) { uint32_t param, val; /* ask for encapsulated CPLs */ param = FW_PARAM_PFVF(CPLFW4MSG_ENCAP); val = 1; (void)t4_set_params(sc, sc->mbox, sc->pf, 0, 1, ¶m, &val); return (0); } #undef FW_PARAM_PFVF #undef FW_PARAM_DEV static void t4_set_desc(struct adapter *sc) { char buf[128]; struct adapter_params *p = &sc->params; snprintf(buf, sizeof(buf), "Chelsio %s %sNIC (rev %d), S/N:%s, " "P/N:%s, E/C:%s", p->vpd.id, is_offload(sc) ? "R" : "", chip_rev(sc), p->vpd.sn, p->vpd.pn, p->vpd.ec); device_set_desc_copy(sc->dev, buf); } static void build_medialist(struct port_info *pi, struct ifmedia *media) { int m; PORT_LOCK(pi); ifmedia_removeall(media); m = IFM_ETHER | IFM_FDX; switch(pi->port_type) { case FW_PORT_TYPE_BT_XFI: case FW_PORT_TYPE_BT_XAUI: ifmedia_add(media, m | IFM_10G_T, 0, NULL); /* fall through */ case FW_PORT_TYPE_BT_SGMII: ifmedia_add(media, m | IFM_1000_T, 0, NULL); ifmedia_add(media, m | IFM_100_TX, 0, NULL); ifmedia_add(media, IFM_ETHER | IFM_AUTO, 0, NULL); ifmedia_set(media, IFM_ETHER | IFM_AUTO); break; case FW_PORT_TYPE_CX4: ifmedia_add(media, m | IFM_10G_CX4, 0, NULL); ifmedia_set(media, m | IFM_10G_CX4); break; case FW_PORT_TYPE_QSFP_10G: case FW_PORT_TYPE_SFP: case FW_PORT_TYPE_FIBER_XFI: case FW_PORT_TYPE_FIBER_XAUI: switch (pi->mod_type) { case FW_PORT_MOD_TYPE_LR: ifmedia_add(media, m | IFM_10G_LR, 0, NULL); ifmedia_set(media, m | IFM_10G_LR); break; case FW_PORT_MOD_TYPE_SR: ifmedia_add(media, m | IFM_10G_SR, 0, NULL); ifmedia_set(media, m | IFM_10G_SR); break; case FW_PORT_MOD_TYPE_LRM: ifmedia_add(media, m | IFM_10G_LRM, 0, NULL); ifmedia_set(media, m | IFM_10G_LRM); break; case FW_PORT_MOD_TYPE_TWINAX_PASSIVE: case FW_PORT_MOD_TYPE_TWINAX_ACTIVE: ifmedia_add(media, m | IFM_10G_TWINAX, 0, NULL); ifmedia_set(media, m | IFM_10G_TWINAX); break; case FW_PORT_MOD_TYPE_NONE: m &= ~IFM_FDX; ifmedia_add(media, m | IFM_NONE, 0, NULL); ifmedia_set(media, m | IFM_NONE); break; case FW_PORT_MOD_TYPE_NA: case FW_PORT_MOD_TYPE_ER: default: device_printf(pi->dev, "unknown port_type (%d), mod_type (%d)\n", pi->port_type, pi->mod_type); ifmedia_add(media, m | IFM_UNKNOWN, 0, NULL); ifmedia_set(media, m | IFM_UNKNOWN); break; } break; case FW_PORT_TYPE_QSFP: switch (pi->mod_type) { case FW_PORT_MOD_TYPE_LR: ifmedia_add(media, m | IFM_40G_LR4, 0, NULL); ifmedia_set(media, m | IFM_40G_LR4); break; case FW_PORT_MOD_TYPE_SR: ifmedia_add(media, m | IFM_40G_SR4, 0, NULL); ifmedia_set(media, m | IFM_40G_SR4); break; case FW_PORT_MOD_TYPE_TWINAX_PASSIVE: case FW_PORT_MOD_TYPE_TWINAX_ACTIVE: ifmedia_add(media, m | IFM_40G_CR4, 0, NULL); ifmedia_set(media, m | IFM_40G_CR4); break; case FW_PORT_MOD_TYPE_NONE: m &= ~IFM_FDX; ifmedia_add(media, m | IFM_NONE, 0, NULL); ifmedia_set(media, m | IFM_NONE); break; default: device_printf(pi->dev, "unknown port_type (%d), mod_type (%d)\n", pi->port_type, pi->mod_type); ifmedia_add(media, m | IFM_UNKNOWN, 0, NULL); ifmedia_set(media, m | IFM_UNKNOWN); break; } break; default: device_printf(pi->dev, "unknown port_type (%d), mod_type (%d)\n", pi->port_type, pi->mod_type); ifmedia_add(media, m | IFM_UNKNOWN, 0, NULL); ifmedia_set(media, m | IFM_UNKNOWN); break; } PORT_UNLOCK(pi); } #define FW_MAC_EXACT_CHUNK 7 /* * Program the port's XGMAC based on parameters in ifnet. The caller also * indicates which parameters should be programmed (the rest are left alone). */ int update_mac_settings(struct ifnet *ifp, int flags) { int rc = 0; struct port_info *pi = ifp->if_softc; struct adapter *sc = pi->adapter; int mtu = -1, promisc = -1, allmulti = -1, vlanex = -1; uint16_t viid = 0xffff; int16_t *xact_addr_filt = NULL; ASSERT_SYNCHRONIZED_OP(sc); KASSERT(flags, ("%s: not told what to update.", __func__)); if (ifp == pi->ifp) { viid = pi->viid; xact_addr_filt = &pi->xact_addr_filt; } #ifdef DEV_NETMAP else if (ifp == pi->nm_ifp) { viid = pi->nm_viid; xact_addr_filt = &pi->nm_xact_addr_filt; } #endif if (flags & XGMAC_MTU) mtu = ifp->if_mtu; if (flags & XGMAC_PROMISC) promisc = ifp->if_flags & IFF_PROMISC ? 1 : 0; if (flags & XGMAC_ALLMULTI) allmulti = ifp->if_flags & IFF_ALLMULTI ? 1 : 0; if (flags & XGMAC_VLANEX) vlanex = ifp->if_capenable & IFCAP_VLAN_HWTAGGING ? 1 : 0; if (flags & (XGMAC_MTU|XGMAC_PROMISC|XGMAC_ALLMULTI|XGMAC_VLANEX)) { rc = -t4_set_rxmode(sc, sc->mbox, viid, mtu, promisc, allmulti, 1, vlanex, false); if (rc) { if_printf(ifp, "set_rxmode (%x) failed: %d\n", flags, rc); return (rc); } } if (flags & XGMAC_UCADDR) { uint8_t ucaddr[ETHER_ADDR_LEN]; bcopy(IF_LLADDR(ifp), ucaddr, sizeof(ucaddr)); rc = t4_change_mac(sc, sc->mbox, viid, *xact_addr_filt, ucaddr, true, true); if (rc < 0) { rc = -rc; if_printf(ifp, "change_mac failed: %d\n", rc); return (rc); } else { *xact_addr_filt = rc; rc = 0; } } if (flags & XGMAC_MCADDRS) { const uint8_t *mcaddr[FW_MAC_EXACT_CHUNK]; int del = 1; uint64_t hash = 0; struct ifmultiaddr *ifma; int i = 0, j; if_maddr_rlock(ifp); TAILQ_FOREACH(ifma, &ifp->if_multiaddrs, ifma_link) { if (ifma->ifma_addr->sa_family != AF_LINK) continue; mcaddr[i] = LLADDR((struct sockaddr_dl *)ifma->ifma_addr); MPASS(ETHER_IS_MULTICAST(mcaddr[i])); i++; if (i == FW_MAC_EXACT_CHUNK) { rc = t4_alloc_mac_filt(sc, sc->mbox, viid, del, i, mcaddr, NULL, &hash, 0); if (rc < 0) { rc = -rc; for (j = 0; j < i; j++) { if_printf(ifp, "failed to add mc address" " %02x:%02x:%02x:" "%02x:%02x:%02x rc=%d\n", mcaddr[j][0], mcaddr[j][1], mcaddr[j][2], mcaddr[j][3], mcaddr[j][4], mcaddr[j][5], rc); } goto mcfail; } del = 0; i = 0; } } if (i > 0) { rc = t4_alloc_mac_filt(sc, sc->mbox, viid, del, i, mcaddr, NULL, &hash, 0); if (rc < 0) { rc = -rc; for (j = 0; j < i; j++) { if_printf(ifp, "failed to add mc address" " %02x:%02x:%02x:" "%02x:%02x:%02x rc=%d\n", mcaddr[j][0], mcaddr[j][1], mcaddr[j][2], mcaddr[j][3], mcaddr[j][4], mcaddr[j][5], rc); } goto mcfail; } } rc = -t4_set_addr_hash(sc, sc->mbox, viid, 0, hash, 0); if (rc != 0) if_printf(ifp, "failed to set mc address hash: %d", rc); mcfail: if_maddr_runlock(ifp); } return (rc); } /* * {begin|end}_synchronized_op must be called from the same thread. */ int begin_synchronized_op(struct adapter *sc, struct port_info *pi, int flags, char *wmesg) { int rc, pri; #ifdef WITNESS /* the caller thinks it's ok to sleep, but is it really? */ if (flags & SLEEP_OK) pause("t4slptst", 1); #endif if (INTR_OK) pri = PCATCH; else pri = 0; ADAPTER_LOCK(sc); for (;;) { if (pi && IS_DOOMED(pi)) { rc = ENXIO; goto done; } if (!IS_BUSY(sc)) { rc = 0; break; } if (!(flags & SLEEP_OK)) { rc = EBUSY; goto done; } if (mtx_sleep(&sc->flags, &sc->sc_lock, pri, wmesg, 0)) { rc = EINTR; goto done; } } KASSERT(!IS_BUSY(sc), ("%s: controller busy.", __func__)); SET_BUSY(sc); #ifdef INVARIANTS sc->last_op = wmesg; sc->last_op_thr = curthread; #endif done: if (!(flags & HOLD_LOCK) || rc) ADAPTER_UNLOCK(sc); return (rc); } /* * {begin|end}_synchronized_op must be called from the same thread. */ void end_synchronized_op(struct adapter *sc, int flags) { if (flags & LOCK_HELD) ADAPTER_LOCK_ASSERT_OWNED(sc); else ADAPTER_LOCK(sc); KASSERT(IS_BUSY(sc), ("%s: controller not busy.", __func__)); CLR_BUSY(sc); wakeup(&sc->flags); ADAPTER_UNLOCK(sc); } static int cxgbe_init_synchronized(struct port_info *pi) { struct adapter *sc = pi->adapter; struct ifnet *ifp = pi->ifp; int rc = 0, i; struct sge_txq *txq; ASSERT_SYNCHRONIZED_OP(sc); if (isset(&sc->open_device_map, pi->port_id)) { KASSERT(ifp->if_drv_flags & IFF_DRV_RUNNING, ("mismatch between open_device_map and if_drv_flags")); return (0); /* already running */ } if (!(sc->flags & FULL_INIT_DONE) && ((rc = adapter_full_init(sc)) != 0)) return (rc); /* error message displayed already */ if (!(pi->flags & PORT_INIT_DONE) && ((rc = port_full_init(pi)) != 0)) return (rc); /* error message displayed already */ rc = update_mac_settings(ifp, XGMAC_ALL); if (rc) goto done; /* error message displayed already */ rc = -t4_enable_vi(sc, sc->mbox, pi->viid, true, true); if (rc != 0) { if_printf(ifp, "enable_vi failed: %d\n", rc); goto done; } /* * Can't fail from this point onwards. Review cxgbe_uninit_synchronized * if this changes. */ for_each_txq(pi, i, txq) { TXQ_LOCK(txq); txq->eq.flags |= EQ_ENABLED; TXQ_UNLOCK(txq); } /* * The first iq of the first port to come up is used for tracing. */ if (sc->traceq < 0) { sc->traceq = sc->sge.rxq[pi->first_rxq].iq.abs_id; t4_write_reg(sc, is_t4(sc) ? A_MPS_TRC_RSS_CONTROL : A_MPS_T5_TRC_RSS_CONTROL, V_RSSCONTROL(pi->tx_chan) | V_QUEUENUMBER(sc->traceq)); pi->flags |= HAS_TRACEQ; } /* all ok */ setbit(&sc->open_device_map, pi->port_id); PORT_LOCK(pi); ifp->if_drv_flags |= IFF_DRV_RUNNING; PORT_UNLOCK(pi); callout_reset(&pi->tick, hz, cxgbe_tick, pi); done: if (rc != 0) cxgbe_uninit_synchronized(pi); return (rc); } /* * Idempotent. */ static int cxgbe_uninit_synchronized(struct port_info *pi) { struct adapter *sc = pi->adapter; struct ifnet *ifp = pi->ifp; int rc, i; struct sge_txq *txq; ASSERT_SYNCHRONIZED_OP(sc); if (!(pi->flags & PORT_INIT_DONE)) { KASSERT(!(ifp->if_drv_flags & IFF_DRV_RUNNING), ("uninited port is running")); return (0); } /* * Disable the VI so that all its data in either direction is discarded * by the MPS. Leave everything else (the queues, interrupts, and 1Hz * tick) intact as the TP can deliver negative advice or data that it's * holding in its RAM (for an offloaded connection) even after the VI is * disabled. */ rc = -t4_enable_vi(sc, sc->mbox, pi->viid, false, false); if (rc) { if_printf(ifp, "disable_vi failed: %d\n", rc); return (rc); } for_each_txq(pi, i, txq) { TXQ_LOCK(txq); txq->eq.flags &= ~EQ_ENABLED; TXQ_UNLOCK(txq); } clrbit(&sc->open_device_map, pi->port_id); PORT_LOCK(pi); ifp->if_drv_flags &= ~IFF_DRV_RUNNING; PORT_UNLOCK(pi); pi->link_cfg.link_ok = 0; pi->link_cfg.speed = 0; pi->linkdnrc = -1; t4_os_link_changed(sc, pi->port_id, 0, -1); return (0); } /* * It is ok for this function to fail midway and return right away. t4_detach * will walk the entire sc->irq list and clean up whatever is valid. */ static int setup_intr_handlers(struct adapter *sc) { int rc, rid, p, q; char s[8]; struct irq *irq; struct port_info *pi; struct sge_rxq *rxq; #ifdef TCP_OFFLOAD struct sge_ofld_rxq *ofld_rxq; #endif #ifdef DEV_NETMAP struct sge_nm_rxq *nm_rxq; #endif /* * Setup interrupts. */ irq = &sc->irq[0]; rid = sc->intr_type == INTR_INTX ? 0 : 1; if (sc->intr_count == 1) return (t4_alloc_irq(sc, irq, rid, t4_intr_all, sc, "all")); /* Multiple interrupts. */ KASSERT(sc->intr_count >= T4_EXTRA_INTR + sc->params.nports, ("%s: too few intr.", __func__)); /* The first one is always error intr */ rc = t4_alloc_irq(sc, irq, rid, t4_intr_err, sc, "err"); if (rc != 0) return (rc); irq++; rid++; /* The second one is always the firmware event queue */ rc = t4_alloc_irq(sc, irq, rid, t4_intr_evt, &sc->sge.fwq, "evt"); if (rc != 0) return (rc); irq++; rid++; for_each_port(sc, p) { pi = sc->port[p]; if (pi->flags & INTR_RXQ) { for_each_rxq(pi, q, rxq) { snprintf(s, sizeof(s), "%d.%d", p, q); rc = t4_alloc_irq(sc, irq, rid, t4_intr, rxq, s); if (rc != 0) return (rc); irq++; rid++; } } #ifdef TCP_OFFLOAD if (pi->flags & INTR_OFLD_RXQ) { for_each_ofld_rxq(pi, q, ofld_rxq) { snprintf(s, sizeof(s), "%d,%d", p, q); rc = t4_alloc_irq(sc, irq, rid, t4_intr, ofld_rxq, s); if (rc != 0) return (rc); irq++; rid++; } } #endif #ifdef DEV_NETMAP if (pi->flags & INTR_NM_RXQ) { for_each_nm_rxq(pi, q, nm_rxq) { snprintf(s, sizeof(s), "%d-%d", p, q); rc = t4_alloc_irq(sc, irq, rid, t4_nm_intr, nm_rxq, s); if (rc != 0) return (rc); irq++; rid++; } } #endif } MPASS(irq == &sc->irq[sc->intr_count]); return (0); } int adapter_full_init(struct adapter *sc) { int rc, i; ASSERT_SYNCHRONIZED_OP(sc); ADAPTER_LOCK_ASSERT_NOTOWNED(sc); KASSERT((sc->flags & FULL_INIT_DONE) == 0, ("%s: FULL_INIT_DONE already", __func__)); /* * queues that belong to the adapter (not any particular port). */ rc = t4_setup_adapter_queues(sc); if (rc != 0) goto done; for (i = 0; i < nitems(sc->tq); i++) { sc->tq[i] = taskqueue_create("t4 taskq", M_NOWAIT, taskqueue_thread_enqueue, &sc->tq[i]); if (sc->tq[i] == NULL) { device_printf(sc->dev, "failed to allocate task queue %d\n", i); rc = ENOMEM; goto done; } taskqueue_start_threads(&sc->tq[i], 1, PI_NET, "%s tq%d", device_get_nameunit(sc->dev), i); } t4_intr_enable(sc); sc->flags |= FULL_INIT_DONE; done: if (rc != 0) adapter_full_uninit(sc); return (rc); } int adapter_full_uninit(struct adapter *sc) { int i; ADAPTER_LOCK_ASSERT_NOTOWNED(sc); t4_teardown_adapter_queues(sc); for (i = 0; i < nitems(sc->tq) && sc->tq[i]; i++) { taskqueue_free(sc->tq[i]); sc->tq[i] = NULL; } sc->flags &= ~FULL_INIT_DONE; return (0); } int port_full_init(struct port_info *pi) { struct adapter *sc = pi->adapter; struct ifnet *ifp = pi->ifp; uint16_t *rss; struct sge_rxq *rxq; int rc, i, j; ASSERT_SYNCHRONIZED_OP(sc); KASSERT((pi->flags & PORT_INIT_DONE) == 0, ("%s: PORT_INIT_DONE already", __func__)); sysctl_ctx_init(&pi->ctx); pi->flags |= PORT_SYSCTL_CTX; /* * Allocate tx/rx/fl queues for this port. */ rc = t4_setup_port_queues(pi); if (rc != 0) goto done; /* error message displayed already */ /* * Setup RSS for this port. Save a copy of the RSS table for later use. */ rss = malloc(pi->rss_size * sizeof (*rss), M_CXGBE, M_ZERO | M_WAITOK); for (i = 0; i < pi->rss_size;) { for_each_rxq(pi, j, rxq) { rss[i++] = rxq->iq.abs_id; if (i == pi->rss_size) break; } } rc = -t4_config_rss_range(sc, sc->mbox, pi->viid, 0, pi->rss_size, rss, pi->rss_size); if (rc != 0) { if_printf(ifp, "rss_config failed: %d\n", rc); goto done; } pi->rss = rss; pi->flags |= PORT_INIT_DONE; done: if (rc != 0) port_full_uninit(pi); return (rc); } /* * Idempotent. */ int port_full_uninit(struct port_info *pi) { struct adapter *sc = pi->adapter; int i; struct sge_rxq *rxq; struct sge_txq *txq; #ifdef TCP_OFFLOAD struct sge_ofld_rxq *ofld_rxq; struct sge_wrq *ofld_txq; #endif if (pi->flags & PORT_INIT_DONE) { /* Need to quiesce queues. */ quiesce_wrq(sc, &sc->sge.ctrlq[pi->port_id]); for_each_txq(pi, i, txq) { quiesce_txq(sc, txq); } #ifdef TCP_OFFLOAD for_each_ofld_txq(pi, i, ofld_txq) { quiesce_wrq(sc, ofld_txq); } #endif for_each_rxq(pi, i, rxq) { quiesce_iq(sc, &rxq->iq); quiesce_fl(sc, &rxq->fl); } #ifdef TCP_OFFLOAD for_each_ofld_rxq(pi, i, ofld_rxq) { quiesce_iq(sc, &ofld_rxq->iq); quiesce_fl(sc, &ofld_rxq->fl); } #endif free(pi->rss, M_CXGBE); } t4_teardown_port_queues(pi); pi->flags &= ~PORT_INIT_DONE; return (0); } static void quiesce_txq(struct adapter *sc, struct sge_txq *txq) { struct sge_eq *eq = &txq->eq; struct sge_qstat *spg = (void *)&eq->desc[eq->sidx]; (void) sc; /* unused */ #ifdef INVARIANTS TXQ_LOCK(txq); MPASS((eq->flags & EQ_ENABLED) == 0); TXQ_UNLOCK(txq); #endif /* Wait for the mp_ring to empty. */ while (!mp_ring_is_idle(txq->r)) { mp_ring_check_drainage(txq->r, 0); pause("rquiesce", 1); } /* Then wait for the hardware to finish. */ while (spg->cidx != htobe16(eq->pidx)) pause("equiesce", 1); /* Finally, wait for the driver to reclaim all descriptors. */ while (eq->cidx != eq->pidx) pause("dquiesce", 1); } static void quiesce_wrq(struct adapter *sc, struct sge_wrq *wrq) { /* XXXTX */ } static void quiesce_iq(struct adapter *sc, struct sge_iq *iq) { (void) sc; /* unused */ /* Synchronize with the interrupt handler */ while (!atomic_cmpset_int(&iq->state, IQS_IDLE, IQS_DISABLED)) pause("iqfree", 1); } static void quiesce_fl(struct adapter *sc, struct sge_fl *fl) { mtx_lock(&sc->sfl_lock); FL_LOCK(fl); fl->flags |= FL_DOOMED; FL_UNLOCK(fl); mtx_unlock(&sc->sfl_lock); callout_drain(&sc->sfl_callout); KASSERT((fl->flags & FL_STARVING) == 0, ("%s: still starving", __func__)); } static int t4_alloc_irq(struct adapter *sc, struct irq *irq, int rid, driver_intr_t *handler, void *arg, char *name) { int rc; irq->rid = rid; irq->res = bus_alloc_resource_any(sc->dev, SYS_RES_IRQ, &irq->rid, RF_SHAREABLE | RF_ACTIVE); if (irq->res == NULL) { device_printf(sc->dev, "failed to allocate IRQ for rid %d, name %s.\n", rid, name); return (ENOMEM); } rc = bus_setup_intr(sc->dev, irq->res, INTR_MPSAFE | INTR_TYPE_NET, NULL, handler, arg, &irq->tag); if (rc != 0) { device_printf(sc->dev, "failed to setup interrupt for rid %d, name %s: %d\n", rid, name, rc); } else if (name) bus_describe_intr(sc->dev, irq->res, irq->tag, name); return (rc); } static int t4_free_irq(struct adapter *sc, struct irq *irq) { if (irq->tag) bus_teardown_intr(sc->dev, irq->res, irq->tag); if (irq->res) bus_release_resource(sc->dev, SYS_RES_IRQ, irq->rid, irq->res); bzero(irq, sizeof(*irq)); return (0); } static void reg_block_dump(struct adapter *sc, uint8_t *buf, unsigned int start, unsigned int end) { uint32_t *p = (uint32_t *)(buf + start); for ( ; start <= end; start += sizeof(uint32_t)) *p++ = t4_read_reg(sc, start); } static void t4_get_regs(struct adapter *sc, struct t4_regdump *regs, uint8_t *buf) { int i, n; const unsigned int *reg_ranges; static const unsigned int t4_reg_ranges[] = { 0x1008, 0x1108, 0x1180, 0x11b4, 0x11fc, 0x123c, 0x1300, 0x173c, 0x1800, 0x18fc, 0x3000, 0x30d8, 0x30e0, 0x5924, 0x5960, 0x59d4, 0x5a00, 0x5af8, 0x6000, 0x6098, 0x6100, 0x6150, 0x6200, 0x6208, 0x6240, 0x6248, 0x6280, 0x6338, 0x6370, 0x638c, 0x6400, 0x643c, 0x6500, 0x6524, 0x6a00, 0x6a38, 0x6a60, 0x6a78, 0x6b00, 0x6b84, 0x6bf0, 0x6c84, 0x6cf0, 0x6d84, 0x6df0, 0x6e84, 0x6ef0, 0x6f84, 0x6ff0, 0x7084, 0x70f0, 0x7184, 0x71f0, 0x7284, 0x72f0, 0x7384, 0x73f0, 0x7450, 0x7500, 0x7530, 0x7600, 0x761c, 0x7680, 0x76cc, 0x7700, 0x7798, 0x77c0, 0x77fc, 0x7900, 0x79fc, 0x7b00, 0x7c38, 0x7d00, 0x7efc, 0x8dc0, 0x8e1c, 0x8e30, 0x8e78, 0x8ea0, 0x8f6c, 0x8fc0, 0x9074, 0x90fc, 0x90fc, 0x9400, 0x9458, 0x9600, 0x96bc, 0x9800, 0x9808, 0x9820, 0x983c, 0x9850, 0x9864, 0x9c00, 0x9c6c, 0x9c80, 0x9cec, 0x9d00, 0x9d6c, 0x9d80, 0x9dec, 0x9e00, 0x9e6c, 0x9e80, 0x9eec, 0x9f00, 0x9f6c, 0x9f80, 0x9fec, 0xd004, 0xd03c, 0xdfc0, 0xdfe0, 0xe000, 0xea7c, 0xf000, 0x11110, 0x11118, 0x11190, 0x19040, 0x1906c, 0x19078, 0x19080, 0x1908c, 0x19124, 0x19150, 0x191b0, 0x191d0, 0x191e8, 0x19238, 0x1924c, 0x193f8, 0x19474, 0x19490, 0x194f8, 0x19800, 0x19f30, 0x1a000, 0x1a06c, 0x1a0b0, 0x1a120, 0x1a128, 0x1a138, 0x1a190, 0x1a1c4, 0x1a1fc, 0x1a1fc, 0x1e040, 0x1e04c, 0x1e284, 0x1e28c, 0x1e2c0, 0x1e2c0, 0x1e2e0, 0x1e2e0, 0x1e300, 0x1e384, 0x1e3c0, 0x1e3c8, 0x1e440, 0x1e44c, 0x1e684, 0x1e68c, 0x1e6c0, 0x1e6c0, 0x1e6e0, 0x1e6e0, 0x1e700, 0x1e784, 0x1e7c0, 0x1e7c8, 0x1e840, 0x1e84c, 0x1ea84, 0x1ea8c, 0x1eac0, 0x1eac0, 0x1eae0, 0x1eae0, 0x1eb00, 0x1eb84, 0x1ebc0, 0x1ebc8, 0x1ec40, 0x1ec4c, 0x1ee84, 0x1ee8c, 0x1eec0, 0x1eec0, 0x1eee0, 0x1eee0, 0x1ef00, 0x1ef84, 0x1efc0, 0x1efc8, 0x1f040, 0x1f04c, 0x1f284, 0x1f28c, 0x1f2c0, 0x1f2c0, 0x1f2e0, 0x1f2e0, 0x1f300, 0x1f384, 0x1f3c0, 0x1f3c8, 0x1f440, 0x1f44c, 0x1f684, 0x1f68c, 0x1f6c0, 0x1f6c0, 0x1f6e0, 0x1f6e0, 0x1f700, 0x1f784, 0x1f7c0, 0x1f7c8, 0x1f840, 0x1f84c, 0x1fa84, 0x1fa8c, 0x1fac0, 0x1fac0, 0x1fae0, 0x1fae0, 0x1fb00, 0x1fb84, 0x1fbc0, 0x1fbc8, 0x1fc40, 0x1fc4c, 0x1fe84, 0x1fe8c, 0x1fec0, 0x1fec0, 0x1fee0, 0x1fee0, 0x1ff00, 0x1ff84, 0x1ffc0, 0x1ffc8, 0x20000, 0x2002c, 0x20100, 0x2013c, 0x20190, 0x201c8, 0x20200, 0x20318, 0x20400, 0x20528, 0x20540, 0x20614, 0x21000, 0x21040, 0x2104c, 0x21060, 0x210c0, 0x210ec, 0x21200, 0x21268, 0x21270, 0x21284, 0x212fc, 0x21388, 0x21400, 0x21404, 0x21500, 0x21518, 0x2152c, 0x2153c, 0x21550, 0x21554, 0x21600, 0x21600, 0x21608, 0x21628, 0x21630, 0x2163c, 0x21700, 0x2171c, 0x21780, 0x2178c, 0x21800, 0x21c38, 0x21c80, 0x21d7c, 0x21e00, 0x21e04, 0x22000, 0x2202c, 0x22100, 0x2213c, 0x22190, 0x221c8, 0x22200, 0x22318, 0x22400, 0x22528, 0x22540, 0x22614, 0x23000, 0x23040, 0x2304c, 0x23060, 0x230c0, 0x230ec, 0x23200, 0x23268, 0x23270, 0x23284, 0x232fc, 0x23388, 0x23400, 0x23404, 0x23500, 0x23518, 0x2352c, 0x2353c, 0x23550, 0x23554, 0x23600, 0x23600, 0x23608, 0x23628, 0x23630, 0x2363c, 0x23700, 0x2371c, 0x23780, 0x2378c, 0x23800, 0x23c38, 0x23c80, 0x23d7c, 0x23e00, 0x23e04, 0x24000, 0x2402c, 0x24100, 0x2413c, 0x24190, 0x241c8, 0x24200, 0x24318, 0x24400, 0x24528, 0x24540, 0x24614, 0x25000, 0x25040, 0x2504c, 0x25060, 0x250c0, 0x250ec, 0x25200, 0x25268, 0x25270, 0x25284, 0x252fc, 0x25388, 0x25400, 0x25404, 0x25500, 0x25518, 0x2552c, 0x2553c, 0x25550, 0x25554, 0x25600, 0x25600, 0x25608, 0x25628, 0x25630, 0x2563c, 0x25700, 0x2571c, 0x25780, 0x2578c, 0x25800, 0x25c38, 0x25c80, 0x25d7c, 0x25e00, 0x25e04, 0x26000, 0x2602c, 0x26100, 0x2613c, 0x26190, 0x261c8, 0x26200, 0x26318, 0x26400, 0x26528, 0x26540, 0x26614, 0x27000, 0x27040, 0x2704c, 0x27060, 0x270c0, 0x270ec, 0x27200, 0x27268, 0x27270, 0x27284, 0x272fc, 0x27388, 0x27400, 0x27404, 0x27500, 0x27518, 0x2752c, 0x2753c, 0x27550, 0x27554, 0x27600, 0x27600, 0x27608, 0x27628, 0x27630, 0x2763c, 0x27700, 0x2771c, 0x27780, 0x2778c, 0x27800, 0x27c38, 0x27c80, 0x27d7c, 0x27e00, 0x27e04 }; static const unsigned int t5_reg_ranges[] = { 0x1008, 0x1148, 0x1180, 0x11b4, 0x11fc, 0x123c, 0x1280, 0x173c, 0x1800, 0x18fc, 0x3000, 0x3028, 0x3060, 0x30d8, 0x30e0, 0x30fc, 0x3140, 0x357c, 0x35a8, 0x35cc, 0x35ec, 0x35ec, 0x3600, 0x5624, 0x56cc, 0x575c, 0x580c, 0x5814, 0x5890, 0x58bc, 0x5940, 0x59dc, 0x59fc, 0x5a18, 0x5a60, 0x5a9c, 0x5b94, 0x5bfc, 0x6000, 0x6040, 0x6058, 0x614c, 0x7700, 0x7798, 0x77c0, 0x78fc, 0x7b00, 0x7c54, 0x7d00, 0x7efc, 0x8dc0, 0x8de0, 0x8df8, 0x8e84, 0x8ea0, 0x8f84, 0x8fc0, 0x90f8, 0x9400, 0x9470, 0x9600, 0x96f4, 0x9800, 0x9808, 0x9820, 0x983c, 0x9850, 0x9864, 0x9c00, 0x9c6c, 0x9c80, 0x9cec, 0x9d00, 0x9d6c, 0x9d80, 0x9dec, 0x9e00, 0x9e6c, 0x9e80, 0x9eec, 0x9f00, 0x9f6c, 0x9f80, 0xa020, 0xd004, 0xd03c, 0xdfc0, 0xdfe0, 0xe000, 0x11088, 0x1109c, 0x11110, 0x11118, 0x1117c, 0x11190, 0x11204, 0x19040, 0x1906c, 0x19078, 0x19080, 0x1908c, 0x19124, 0x19150, 0x191b0, 0x191d0, 0x191e8, 0x19238, 0x19290, 0x193f8, 0x19474, 0x19490, 0x194cc, 0x194f0, 0x194f8, 0x19c00, 0x19c60, 0x19c94, 0x19e10, 0x19e50, 0x19f34, 0x19f40, 0x19f50, 0x19f90, 0x19fe4, 0x1a000, 0x1a06c, 0x1a0b0, 0x1a120, 0x1a128, 0x1a138, 0x1a190, 0x1a1c4, 0x1a1fc, 0x1a1fc, 0x1e008, 0x1e00c, 0x1e040, 0x1e04c, 0x1e284, 0x1e290, 0x1e2c0, 0x1e2c0, 0x1e2e0, 0x1e2e0, 0x1e300, 0x1e384, 0x1e3c0, 0x1e3c8, 0x1e408, 0x1e40c, 0x1e440, 0x1e44c, 0x1e684, 0x1e690, 0x1e6c0, 0x1e6c0, 0x1e6e0, 0x1e6e0, 0x1e700, 0x1e784, 0x1e7c0, 0x1e7c8, 0x1e808, 0x1e80c, 0x1e840, 0x1e84c, 0x1ea84, 0x1ea90, 0x1eac0, 0x1eac0, 0x1eae0, 0x1eae0, 0x1eb00, 0x1eb84, 0x1ebc0, 0x1ebc8, 0x1ec08, 0x1ec0c, 0x1ec40, 0x1ec4c, 0x1ee84, 0x1ee90, 0x1eec0, 0x1eec0, 0x1eee0, 0x1eee0, 0x1ef00, 0x1ef84, 0x1efc0, 0x1efc8, 0x1f008, 0x1f00c, 0x1f040, 0x1f04c, 0x1f284, 0x1f290, 0x1f2c0, 0x1f2c0, 0x1f2e0, 0x1f2e0, 0x1f300, 0x1f384, 0x1f3c0, 0x1f3c8, 0x1f408, 0x1f40c, 0x1f440, 0x1f44c, 0x1f684, 0x1f690, 0x1f6c0, 0x1f6c0, 0x1f6e0, 0x1f6e0, 0x1f700, 0x1f784, 0x1f7c0, 0x1f7c8, 0x1f808, 0x1f80c, 0x1f840, 0x1f84c, 0x1fa84, 0x1fa90, 0x1fac0, 0x1fac0, 0x1fae0, 0x1fae0, 0x1fb00, 0x1fb84, 0x1fbc0, 0x1fbc8, 0x1fc08, 0x1fc0c, 0x1fc40, 0x1fc4c, 0x1fe84, 0x1fe90, 0x1fec0, 0x1fec0, 0x1fee0, 0x1fee0, 0x1ff00, 0x1ff84, 0x1ffc0, 0x1ffc8, 0x30000, 0x30030, 0x30100, 0x30144, 0x30190, 0x301d0, 0x30200, 0x30318, 0x30400, 0x3052c, 0x30540, 0x3061c, 0x30800, 0x30834, 0x308c0, 0x30908, 0x30910, 0x309ac, 0x30a00, 0x30a2c, 0x30a44, 0x30a50, 0x30a74, 0x30c24, 0x30d00, 0x30d00, 0x30d08, 0x30d14, 0x30d1c, 0x30d20, 0x30d3c, 0x30d50, 0x31200, 0x3120c, 0x31220, 0x31220, 0x31240, 0x31240, 0x31600, 0x3160c, 0x31a00, 0x31a1c, 0x31e00, 0x31e20, 0x31e38, 0x31e3c, 0x31e80, 0x31e80, 0x31e88, 0x31ea8, 0x31eb0, 0x31eb4, 0x31ec8, 0x31ed4, 0x31fb8, 0x32004, 0x32200, 0x32200, 0x32208, 0x32240, 0x32248, 0x32280, 0x32288, 0x322c0, 0x322c8, 0x322fc, 0x32600, 0x32630, 0x32a00, 0x32abc, 0x32b00, 0x32b70, 0x33000, 0x33048, 0x33060, 0x3309c, 0x330f0, 0x33148, 0x33160, 0x3319c, 0x331f0, 0x332e4, 0x332f8, 0x333e4, 0x333f8, 0x33448, 0x33460, 0x3349c, 0x334f0, 0x33548, 0x33560, 0x3359c, 0x335f0, 0x336e4, 0x336f8, 0x337e4, 0x337f8, 0x337fc, 0x33814, 0x33814, 0x3382c, 0x3382c, 0x33880, 0x3388c, 0x338e8, 0x338ec, 0x33900, 0x33948, 0x33960, 0x3399c, 0x339f0, 0x33ae4, 0x33af8, 0x33b10, 0x33b28, 0x33b28, 0x33b3c, 0x33b50, 0x33bf0, 0x33c10, 0x33c28, 0x33c28, 0x33c3c, 0x33c50, 0x33cf0, 0x33cfc, 0x34000, 0x34030, 0x34100, 0x34144, 0x34190, 0x341d0, 0x34200, 0x34318, 0x34400, 0x3452c, 0x34540, 0x3461c, 0x34800, 0x34834, 0x348c0, 0x34908, 0x34910, 0x349ac, 0x34a00, 0x34a2c, 0x34a44, 0x34a50, 0x34a74, 0x34c24, 0x34d00, 0x34d00, 0x34d08, 0x34d14, 0x34d1c, 0x34d20, 0x34d3c, 0x34d50, 0x35200, 0x3520c, 0x35220, 0x35220, 0x35240, 0x35240, 0x35600, 0x3560c, 0x35a00, 0x35a1c, 0x35e00, 0x35e20, 0x35e38, 0x35e3c, 0x35e80, 0x35e80, 0x35e88, 0x35ea8, 0x35eb0, 0x35eb4, 0x35ec8, 0x35ed4, 0x35fb8, 0x36004, 0x36200, 0x36200, 0x36208, 0x36240, 0x36248, 0x36280, 0x36288, 0x362c0, 0x362c8, 0x362fc, 0x36600, 0x36630, 0x36a00, 0x36abc, 0x36b00, 0x36b70, 0x37000, 0x37048, 0x37060, 0x3709c, 0x370f0, 0x37148, 0x37160, 0x3719c, 0x371f0, 0x372e4, 0x372f8, 0x373e4, 0x373f8, 0x37448, 0x37460, 0x3749c, 0x374f0, 0x37548, 0x37560, 0x3759c, 0x375f0, 0x376e4, 0x376f8, 0x377e4, 0x377f8, 0x377fc, 0x37814, 0x37814, 0x3782c, 0x3782c, 0x37880, 0x3788c, 0x378e8, 0x378ec, 0x37900, 0x37948, 0x37960, 0x3799c, 0x379f0, 0x37ae4, 0x37af8, 0x37b10, 0x37b28, 0x37b28, 0x37b3c, 0x37b50, 0x37bf0, 0x37c10, 0x37c28, 0x37c28, 0x37c3c, 0x37c50, 0x37cf0, 0x37cfc, 0x38000, 0x38030, 0x38100, 0x38144, 0x38190, 0x381d0, 0x38200, 0x38318, 0x38400, 0x3852c, 0x38540, 0x3861c, 0x38800, 0x38834, 0x388c0, 0x38908, 0x38910, 0x389ac, 0x38a00, 0x38a2c, 0x38a44, 0x38a50, 0x38a74, 0x38c24, 0x38d00, 0x38d00, 0x38d08, 0x38d14, 0x38d1c, 0x38d20, 0x38d3c, 0x38d50, 0x39200, 0x3920c, 0x39220, 0x39220, 0x39240, 0x39240, 0x39600, 0x3960c, 0x39a00, 0x39a1c, 0x39e00, 0x39e20, 0x39e38, 0x39e3c, 0x39e80, 0x39e80, 0x39e88, 0x39ea8, 0x39eb0, 0x39eb4, 0x39ec8, 0x39ed4, 0x39fb8, 0x3a004, 0x3a200, 0x3a200, 0x3a208, 0x3a240, 0x3a248, 0x3a280, 0x3a288, 0x3a2c0, 0x3a2c8, 0x3a2fc, 0x3a600, 0x3a630, 0x3aa00, 0x3aabc, 0x3ab00, 0x3ab70, 0x3b000, 0x3b048, 0x3b060, 0x3b09c, 0x3b0f0, 0x3b148, 0x3b160, 0x3b19c, 0x3b1f0, 0x3b2e4, 0x3b2f8, 0x3b3e4, 0x3b3f8, 0x3b448, 0x3b460, 0x3b49c, 0x3b4f0, 0x3b548, 0x3b560, 0x3b59c, 0x3b5f0, 0x3b6e4, 0x3b6f8, 0x3b7e4, 0x3b7f8, 0x3b7fc, 0x3b814, 0x3b814, 0x3b82c, 0x3b82c, 0x3b880, 0x3b88c, 0x3b8e8, 0x3b8ec, 0x3b900, 0x3b948, 0x3b960, 0x3b99c, 0x3b9f0, 0x3bae4, 0x3baf8, 0x3bb10, 0x3bb28, 0x3bb28, 0x3bb3c, 0x3bb50, 0x3bbf0, 0x3bc10, 0x3bc28, 0x3bc28, 0x3bc3c, 0x3bc50, 0x3bcf0, 0x3bcfc, 0x3c000, 0x3c030, 0x3c100, 0x3c144, 0x3c190, 0x3c1d0, 0x3c200, 0x3c318, 0x3c400, 0x3c52c, 0x3c540, 0x3c61c, 0x3c800, 0x3c834, 0x3c8c0, 0x3c908, 0x3c910, 0x3c9ac, 0x3ca00, 0x3ca2c, 0x3ca44, 0x3ca50, 0x3ca74, 0x3cc24, 0x3cd00, 0x3cd00, 0x3cd08, 0x3cd14, 0x3cd1c, 0x3cd20, 0x3cd3c, 0x3cd50, 0x3d200, 0x3d20c, 0x3d220, 0x3d220, 0x3d240, 0x3d240, 0x3d600, 0x3d60c, 0x3da00, 0x3da1c, 0x3de00, 0x3de20, 0x3de38, 0x3de3c, 0x3de80, 0x3de80, 0x3de88, 0x3dea8, 0x3deb0, 0x3deb4, 0x3dec8, 0x3ded4, 0x3dfb8, 0x3e004, 0x3e200, 0x3e200, 0x3e208, 0x3e240, 0x3e248, 0x3e280, 0x3e288, 0x3e2c0, 0x3e2c8, 0x3e2fc, 0x3e600, 0x3e630, 0x3ea00, 0x3eabc, 0x3eb00, 0x3eb70, 0x3f000, 0x3f048, 0x3f060, 0x3f09c, 0x3f0f0, 0x3f148, 0x3f160, 0x3f19c, 0x3f1f0, 0x3f2e4, 0x3f2f8, 0x3f3e4, 0x3f3f8, 0x3f448, 0x3f460, 0x3f49c, 0x3f4f0, 0x3f548, 0x3f560, 0x3f59c, 0x3f5f0, 0x3f6e4, 0x3f6f8, 0x3f7e4, 0x3f7f8, 0x3f7fc, 0x3f814, 0x3f814, 0x3f82c, 0x3f82c, 0x3f880, 0x3f88c, 0x3f8e8, 0x3f8ec, 0x3f900, 0x3f948, 0x3f960, 0x3f99c, 0x3f9f0, 0x3fae4, 0x3faf8, 0x3fb10, 0x3fb28, 0x3fb28, 0x3fb3c, 0x3fb50, 0x3fbf0, 0x3fc10, 0x3fc28, 0x3fc28, 0x3fc3c, 0x3fc50, 0x3fcf0, 0x3fcfc, 0x40000, 0x4000c, 0x40040, 0x40068, 0x4007c, 0x40144, 0x40180, 0x4018c, 0x40200, 0x40298, 0x402ac, 0x4033c, 0x403f8, 0x403fc, 0x41304, 0x413c4, 0x41400, 0x4141c, 0x41480, 0x414d0, 0x44000, 0x44078, 0x440c0, 0x44278, 0x442c0, 0x44478, 0x444c0, 0x44678, 0x446c0, 0x44878, 0x448c0, 0x449fc, 0x45000, 0x45068, 0x45080, 0x45084, 0x450a0, 0x450b0, 0x45200, 0x45268, 0x45280, 0x45284, 0x452a0, 0x452b0, 0x460c0, 0x460e4, 0x47000, 0x4708c, 0x47200, 0x47250, 0x47400, 0x47420, 0x47600, 0x47618, 0x47800, 0x47814, 0x48000, 0x4800c, 0x48040, 0x48068, 0x4807c, 0x48144, 0x48180, 0x4818c, 0x48200, 0x48298, 0x482ac, 0x4833c, 0x483f8, 0x483fc, 0x49304, 0x493c4, 0x49400, 0x4941c, 0x49480, 0x494d0, 0x4c000, 0x4c078, 0x4c0c0, 0x4c278, 0x4c2c0, 0x4c478, 0x4c4c0, 0x4c678, 0x4c6c0, 0x4c878, 0x4c8c0, 0x4c9fc, 0x4d000, 0x4d068, 0x4d080, 0x4d084, 0x4d0a0, 0x4d0b0, 0x4d200, 0x4d268, 0x4d280, 0x4d284, 0x4d2a0, 0x4d2b0, 0x4e0c0, 0x4e0e4, 0x4f000, 0x4f08c, 0x4f200, 0x4f250, 0x4f400, 0x4f420, 0x4f600, 0x4f618, 0x4f800, 0x4f814, 0x50000, 0x500cc, 0x50400, 0x50400, 0x50800, 0x508cc, 0x50c00, 0x50c00, 0x51000, 0x5101c, 0x51300, 0x51308, }; if (is_t4(sc)) { reg_ranges = &t4_reg_ranges[0]; n = nitems(t4_reg_ranges); } else { reg_ranges = &t5_reg_ranges[0]; n = nitems(t5_reg_ranges); } regs->version = chip_id(sc) | chip_rev(sc) << 10; for (i = 0; i < n; i += 2) reg_block_dump(sc, buf, reg_ranges[i], reg_ranges[i + 1]); } static void cxgbe_refresh_stats(struct adapter *sc, struct port_info *pi) { int i; u_int v, tnl_cong_drops; struct timeval tv; const struct timeval interval = {0, 250000}; /* 250ms */ getmicrotime(&tv); timevalsub(&tv, &interval); if (timevalcmp(&tv, &pi->last_refreshed, <)) return; tnl_cong_drops = 0; t4_get_port_stats(sc, pi->tx_chan, &pi->stats); for (i = 0; i < NCHAN; i++) { if (pi->rx_chan_map & (1 << i)) { mtx_lock(&sc->regwin_lock); t4_read_indirect(sc, A_TP_MIB_INDEX, A_TP_MIB_DATA, &v, 1, A_TP_MIB_TNL_CNG_DROP_0 + i); mtx_unlock(&sc->regwin_lock); tnl_cong_drops += v; } } pi->tnl_cong_drops = tnl_cong_drops; getmicrotime(&pi->last_refreshed); } static void cxgbe_tick(void *arg) { struct port_info *pi = arg; struct adapter *sc = pi->adapter; struct ifnet *ifp = pi->ifp; PORT_LOCK(pi); if (!(ifp->if_drv_flags & IFF_DRV_RUNNING)) { PORT_UNLOCK(pi); return; /* without scheduling another callout */ } cxgbe_refresh_stats(sc, pi); callout_schedule(&pi->tick, hz); PORT_UNLOCK(pi); } static void cxgbe_vlan_config(void *arg, struct ifnet *ifp, uint16_t vid) { struct ifnet *vlan; if (arg != ifp || ifp->if_type != IFT_ETHER) return; vlan = VLAN_DEVAT(ifp, vid); VLAN_SETCOOKIE(vlan, ifp); } static int cpl_not_handled(struct sge_iq *iq, const struct rss_header *rss, struct mbuf *m) { #ifdef INVARIANTS panic("%s: opcode 0x%02x on iq %p with payload %p", __func__, rss->opcode, iq, m); #else log(LOG_ERR, "%s: opcode 0x%02x on iq %p with payload %p\n", __func__, rss->opcode, iq, m); m_freem(m); #endif return (EDOOFUS); } int t4_register_cpl_handler(struct adapter *sc, int opcode, cpl_handler_t h) { uintptr_t *loc, new; if (opcode >= nitems(sc->cpl_handler)) return (EINVAL); new = h ? (uintptr_t)h : (uintptr_t)cpl_not_handled; loc = (uintptr_t *) &sc->cpl_handler[opcode]; atomic_store_rel_ptr(loc, new); return (0); } static int an_not_handled(struct sge_iq *iq, const struct rsp_ctrl *ctrl) { #ifdef INVARIANTS panic("%s: async notification on iq %p (ctrl %p)", __func__, iq, ctrl); #else log(LOG_ERR, "%s: async notification on iq %p (ctrl %p)\n", __func__, iq, ctrl); #endif return (EDOOFUS); } int t4_register_an_handler(struct adapter *sc, an_handler_t h) { uintptr_t *loc, new; new = h ? (uintptr_t)h : (uintptr_t)an_not_handled; loc = (uintptr_t *) &sc->an_handler; atomic_store_rel_ptr(loc, new); return (0); } static int fw_msg_not_handled(struct adapter *sc, const __be64 *rpl) { const struct cpl_fw6_msg *cpl = __containerof(rpl, struct cpl_fw6_msg, data[0]); #ifdef INVARIANTS panic("%s: fw_msg type %d", __func__, cpl->type); #else log(LOG_ERR, "%s: fw_msg type %d\n", __func__, cpl->type); #endif return (EDOOFUS); } int t4_register_fw_msg_handler(struct adapter *sc, int type, fw_msg_handler_t h) { uintptr_t *loc, new; if (type >= nitems(sc->fw_msg_handler)) return (EINVAL); /* * These are dispatched by the handler for FW{4|6}_CPL_MSG using the CPL * handler dispatch table. Reject any attempt to install a handler for * this subtype. */ if (type == FW_TYPE_RSSCPL || type == FW6_TYPE_RSSCPL) return (EINVAL); new = h ? (uintptr_t)h : (uintptr_t)fw_msg_not_handled; loc = (uintptr_t *) &sc->fw_msg_handler[type]; atomic_store_rel_ptr(loc, new); return (0); } static int t4_sysctls(struct adapter *sc) { struct sysctl_ctx_list *ctx; struct sysctl_oid *oid; struct sysctl_oid_list *children, *c0; static char *caps[] = { "\20\1PPP\2QFC\3DCBX", /* caps[0] linkcaps */ "\20\1NIC\2VM\3IDS\4UM\5UM_ISGL" /* caps[1] niccaps */ "\6HASHFILTER\7ETHOFLD", "\20\1TOE", /* caps[2] toecaps */ "\20\1RDDP\2RDMAC", /* caps[3] rdmacaps */ "\20\1INITIATOR_PDU\2TARGET_PDU" /* caps[4] iscsicaps */ "\3INITIATOR_CNXOFLD\4TARGET_CNXOFLD" "\5INITIATOR_SSNOFLD\6TARGET_SSNOFLD", "\20\1INITIATOR\2TARGET\3CTRL_OFLD" /* caps[5] fcoecaps */ "\4PO_INITIAOR\5PO_TARGET" }; static char *doorbells = {"\20\1UDB\2WCWR\3UDBWC\4KDB"}; ctx = device_get_sysctl_ctx(sc->dev); /* * dev.t4nex.X. */ oid = device_get_sysctl_tree(sc->dev); c0 = children = SYSCTL_CHILDREN(oid); sc->sc_do_rxcopy = 1; SYSCTL_ADD_INT(ctx, children, OID_AUTO, "do_rx_copy", CTLFLAG_RW, &sc->sc_do_rxcopy, 1, "Do RX copy of small frames"); SYSCTL_ADD_INT(ctx, children, OID_AUTO, "nports", CTLFLAG_RD, NULL, sc->params.nports, "# of ports"); SYSCTL_ADD_INT(ctx, children, OID_AUTO, "hw_revision", CTLFLAG_RD, NULL, chip_rev(sc), "chip hardware revision"); SYSCTL_ADD_STRING(ctx, children, OID_AUTO, "firmware_version", CTLFLAG_RD, sc->fw_version, 0, "firmware version"); SYSCTL_ADD_STRING(ctx, children, OID_AUTO, "cf", CTLFLAG_RD, sc->cfg_file, 0, "configuration file"); SYSCTL_ADD_UINT(ctx, children, OID_AUTO, "cfcsum", CTLFLAG_RD, NULL, sc->cfcsum, "config file checksum"); SYSCTL_ADD_PROC(ctx, children, OID_AUTO, "doorbells", CTLTYPE_STRING | CTLFLAG_RD, doorbells, sc->doorbells, sysctl_bitfield, "A", "available doorbells"); SYSCTL_ADD_PROC(ctx, children, OID_AUTO, "linkcaps", CTLTYPE_STRING | CTLFLAG_RD, caps[0], sc->linkcaps, sysctl_bitfield, "A", "available link capabilities"); SYSCTL_ADD_PROC(ctx, children, OID_AUTO, "niccaps", CTLTYPE_STRING | CTLFLAG_RD, caps[1], sc->niccaps, sysctl_bitfield, "A", "available NIC capabilities"); SYSCTL_ADD_PROC(ctx, children, OID_AUTO, "toecaps", CTLTYPE_STRING | CTLFLAG_RD, caps[2], sc->toecaps, sysctl_bitfield, "A", "available TCP offload capabilities"); SYSCTL_ADD_PROC(ctx, children, OID_AUTO, "rdmacaps", CTLTYPE_STRING | CTLFLAG_RD, caps[3], sc->rdmacaps, sysctl_bitfield, "A", "available RDMA capabilities"); SYSCTL_ADD_PROC(ctx, children, OID_AUTO, "iscsicaps", CTLTYPE_STRING | CTLFLAG_RD, caps[4], sc->iscsicaps, sysctl_bitfield, "A", "available iSCSI capabilities"); SYSCTL_ADD_PROC(ctx, children, OID_AUTO, "fcoecaps", CTLTYPE_STRING | CTLFLAG_RD, caps[5], sc->fcoecaps, sysctl_bitfield, "A", "available FCoE capabilities"); SYSCTL_ADD_INT(ctx, children, OID_AUTO, "core_clock", CTLFLAG_RD, NULL, sc->params.vpd.cclk, "core clock frequency (in KHz)"); SYSCTL_ADD_PROC(ctx, children, OID_AUTO, "holdoff_timers", CTLTYPE_STRING | CTLFLAG_RD, sc->sge.timer_val, sizeof(sc->sge.timer_val), sysctl_int_array, "A", "interrupt holdoff timer values (us)"); SYSCTL_ADD_PROC(ctx, children, OID_AUTO, "holdoff_pkt_counts", CTLTYPE_STRING | CTLFLAG_RD, sc->sge.counter_val, sizeof(sc->sge.counter_val), sysctl_int_array, "A", "interrupt holdoff packet counter values"); SYSCTL_ADD_INT(ctx, children, OID_AUTO, "nfilters", CTLFLAG_RD, NULL, sc->tids.nftids, "number of filters"); SYSCTL_ADD_PROC(ctx, children, OID_AUTO, "temperature", CTLTYPE_INT | CTLFLAG_RD, sc, 0, sysctl_temperature, "I", "chip temperature (in Celsius)"); t4_sge_sysctls(sc, ctx, children); sc->lro_timeout = 100; SYSCTL_ADD_INT(ctx, children, OID_AUTO, "lro_timeout", CTLFLAG_RW, &sc->lro_timeout, 0, "lro inactive-flush timeout (in us)"); #ifdef SBUF_DRAIN /* * dev.t4nex.X.misc. Marked CTLFLAG_SKIP to avoid information overload. */ oid = SYSCTL_ADD_NODE(ctx, c0, OID_AUTO, "misc", CTLFLAG_RD | CTLFLAG_SKIP, NULL, "logs and miscellaneous information"); children = SYSCTL_CHILDREN(oid); SYSCTL_ADD_PROC(ctx, children, OID_AUTO, "cctrl", CTLTYPE_STRING | CTLFLAG_RD, sc, 0, sysctl_cctrl, "A", "congestion control"); SYSCTL_ADD_PROC(ctx, children, OID_AUTO, "cim_ibq_tp0", CTLTYPE_STRING | CTLFLAG_RD, sc, 0, sysctl_cim_ibq_obq, "A", "CIM IBQ 0 (TP0)"); SYSCTL_ADD_PROC(ctx, children, OID_AUTO, "cim_ibq_tp1", CTLTYPE_STRING | CTLFLAG_RD, sc, 1, sysctl_cim_ibq_obq, "A", "CIM IBQ 1 (TP1)"); SYSCTL_ADD_PROC(ctx, children, OID_AUTO, "cim_ibq_ulp", CTLTYPE_STRING | CTLFLAG_RD, sc, 2, sysctl_cim_ibq_obq, "A", "CIM IBQ 2 (ULP)"); SYSCTL_ADD_PROC(ctx, children, OID_AUTO, "cim_ibq_sge0", CTLTYPE_STRING | CTLFLAG_RD, sc, 3, sysctl_cim_ibq_obq, "A", "CIM IBQ 3 (SGE0)"); SYSCTL_ADD_PROC(ctx, children, OID_AUTO, "cim_ibq_sge1", CTLTYPE_STRING | CTLFLAG_RD, sc, 4, sysctl_cim_ibq_obq, "A", "CIM IBQ 4 (SGE1)"); SYSCTL_ADD_PROC(ctx, children, OID_AUTO, "cim_ibq_ncsi", CTLTYPE_STRING | CTLFLAG_RD, sc, 5, sysctl_cim_ibq_obq, "A", "CIM IBQ 5 (NCSI)"); SYSCTL_ADD_PROC(ctx, children, OID_AUTO, "cim_la", CTLTYPE_STRING | CTLFLAG_RD, sc, 0, sysctl_cim_la, "A", "CIM logic analyzer"); SYSCTL_ADD_PROC(ctx, children, OID_AUTO, "cim_ma_la", CTLTYPE_STRING | CTLFLAG_RD, sc, 0, sysctl_cim_ma_la, "A", "CIM MA logic analyzer"); SYSCTL_ADD_PROC(ctx, children, OID_AUTO, "cim_obq_ulp0", CTLTYPE_STRING | CTLFLAG_RD, sc, 0 + CIM_NUM_IBQ, sysctl_cim_ibq_obq, "A", "CIM OBQ 0 (ULP0)"); SYSCTL_ADD_PROC(ctx, children, OID_AUTO, "cim_obq_ulp1", CTLTYPE_STRING | CTLFLAG_RD, sc, 1 + CIM_NUM_IBQ, sysctl_cim_ibq_obq, "A", "CIM OBQ 1 (ULP1)"); SYSCTL_ADD_PROC(ctx, children, OID_AUTO, "cim_obq_ulp2", CTLTYPE_STRING | CTLFLAG_RD, sc, 2 + CIM_NUM_IBQ, sysctl_cim_ibq_obq, "A", "CIM OBQ 2 (ULP2)"); SYSCTL_ADD_PROC(ctx, children, OID_AUTO, "cim_obq_ulp3", CTLTYPE_STRING | CTLFLAG_RD, sc, 3 + CIM_NUM_IBQ, sysctl_cim_ibq_obq, "A", "CIM OBQ 3 (ULP3)"); SYSCTL_ADD_PROC(ctx, children, OID_AUTO, "cim_obq_sge", CTLTYPE_STRING | CTLFLAG_RD, sc, 4 + CIM_NUM_IBQ, sysctl_cim_ibq_obq, "A", "CIM OBQ 4 (SGE)"); SYSCTL_ADD_PROC(ctx, children, OID_AUTO, "cim_obq_ncsi", CTLTYPE_STRING | CTLFLAG_RD, sc, 5 + CIM_NUM_IBQ, sysctl_cim_ibq_obq, "A", "CIM OBQ 5 (NCSI)"); if (is_t5(sc)) { SYSCTL_ADD_PROC(ctx, children, OID_AUTO, "cim_obq_sge0_rx", CTLTYPE_STRING | CTLFLAG_RD, sc, 6 + CIM_NUM_IBQ, sysctl_cim_ibq_obq, "A", "CIM OBQ 6 (SGE0-RX)"); SYSCTL_ADD_PROC(ctx, children, OID_AUTO, "cim_obq_sge1_rx", CTLTYPE_STRING | CTLFLAG_RD, sc, 7 + CIM_NUM_IBQ, sysctl_cim_ibq_obq, "A", "CIM OBQ 7 (SGE1-RX)"); } SYSCTL_ADD_PROC(ctx, children, OID_AUTO, "cim_pif_la", CTLTYPE_STRING | CTLFLAG_RD, sc, 0, sysctl_cim_pif_la, "A", "CIM PIF logic analyzer"); SYSCTL_ADD_PROC(ctx, children, OID_AUTO, "cim_qcfg", CTLTYPE_STRING | CTLFLAG_RD, sc, 0, sysctl_cim_qcfg, "A", "CIM queue configuration"); SYSCTL_ADD_PROC(ctx, children, OID_AUTO, "cpl_stats", CTLTYPE_STRING | CTLFLAG_RD, sc, 0, sysctl_cpl_stats, "A", "CPL statistics"); SYSCTL_ADD_PROC(ctx, children, OID_AUTO, "ddp_stats", CTLTYPE_STRING | CTLFLAG_RD, sc, 0, sysctl_ddp_stats, "A", "non-TCP DDP statistics"); SYSCTL_ADD_PROC(ctx, children, OID_AUTO, "devlog", CTLTYPE_STRING | CTLFLAG_RD, sc, 0, sysctl_devlog, "A", "firmware's device log"); SYSCTL_ADD_PROC(ctx, children, OID_AUTO, "fcoe_stats", CTLTYPE_STRING | CTLFLAG_RD, sc, 0, sysctl_fcoe_stats, "A", "FCoE statistics"); SYSCTL_ADD_PROC(ctx, children, OID_AUTO, "hw_sched", CTLTYPE_STRING | CTLFLAG_RD, sc, 0, sysctl_hw_sched, "A", "hardware scheduler "); SYSCTL_ADD_PROC(ctx, children, OID_AUTO, "l2t", CTLTYPE_STRING | CTLFLAG_RD, sc, 0, sysctl_l2t, "A", "hardware L2 table"); SYSCTL_ADD_PROC(ctx, children, OID_AUTO, "lb_stats", CTLTYPE_STRING | CTLFLAG_RD, sc, 0, sysctl_lb_stats, "A", "loopback statistics"); SYSCTL_ADD_PROC(ctx, children, OID_AUTO, "meminfo", CTLTYPE_STRING | CTLFLAG_RD, sc, 0, sysctl_meminfo, "A", "memory regions"); SYSCTL_ADD_PROC(ctx, children, OID_AUTO, "mps_tcam", CTLTYPE_STRING | CTLFLAG_RD, sc, 0, sysctl_mps_tcam, "A", "MPS TCAM entries"); SYSCTL_ADD_PROC(ctx, children, OID_AUTO, "path_mtus", CTLTYPE_STRING | CTLFLAG_RD, sc, 0, sysctl_path_mtus, "A", "path MTUs"); SYSCTL_ADD_PROC(ctx, children, OID_AUTO, "pm_stats", CTLTYPE_STRING | CTLFLAG_RD, sc, 0, sysctl_pm_stats, "A", "PM statistics"); SYSCTL_ADD_PROC(ctx, children, OID_AUTO, "rdma_stats", CTLTYPE_STRING | CTLFLAG_RD, sc, 0, sysctl_rdma_stats, "A", "RDMA statistics"); SYSCTL_ADD_PROC(ctx, children, OID_AUTO, "tcp_stats", CTLTYPE_STRING | CTLFLAG_RD, sc, 0, sysctl_tcp_stats, "A", "TCP statistics"); SYSCTL_ADD_PROC(ctx, children, OID_AUTO, "tids", CTLTYPE_STRING | CTLFLAG_RD, sc, 0, sysctl_tids, "A", "TID information"); SYSCTL_ADD_PROC(ctx, children, OID_AUTO, "tp_err_stats", CTLTYPE_STRING | CTLFLAG_RD, sc, 0, sysctl_tp_err_stats, "A", "TP error statistics"); SYSCTL_ADD_PROC(ctx, children, OID_AUTO, "tp_la", CTLTYPE_STRING | CTLFLAG_RD, sc, 0, sysctl_tp_la, "A", "TP logic analyzer"); SYSCTL_ADD_PROC(ctx, children, OID_AUTO, "tx_rate", CTLTYPE_STRING | CTLFLAG_RD, sc, 0, sysctl_tx_rate, "A", "Tx rate"); SYSCTL_ADD_PROC(ctx, children, OID_AUTO, "ulprx_la", CTLTYPE_STRING | CTLFLAG_RD, sc, 0, sysctl_ulprx_la, "A", "ULPRX logic analyzer"); if (is_t5(sc)) { SYSCTL_ADD_PROC(ctx, children, OID_AUTO, "wcwr_stats", CTLTYPE_STRING | CTLFLAG_RD, sc, 0, sysctl_wcwr_stats, "A", "write combined work requests"); } #endif #ifdef TCP_OFFLOAD if (is_offload(sc)) { /* * dev.t4nex.X.toe. */ oid = SYSCTL_ADD_NODE(ctx, c0, OID_AUTO, "toe", CTLFLAG_RD, NULL, "TOE parameters"); children = SYSCTL_CHILDREN(oid); sc->tt.sndbuf = 256 * 1024; SYSCTL_ADD_INT(ctx, children, OID_AUTO, "sndbuf", CTLFLAG_RW, &sc->tt.sndbuf, 0, "max hardware send buffer size"); sc->tt.ddp = 0; SYSCTL_ADD_INT(ctx, children, OID_AUTO, "ddp", CTLFLAG_RW, &sc->tt.ddp, 0, "DDP allowed"); sc->tt.indsz = G_INDICATESIZE(t4_read_reg(sc, A_TP_PARA_REG5)); SYSCTL_ADD_INT(ctx, children, OID_AUTO, "indsz", CTLFLAG_RW, &sc->tt.indsz, 0, "DDP max indicate size allowed"); sc->tt.ddp_thres = G_RXCOALESCESIZE(t4_read_reg(sc, A_TP_PARA_REG2)); SYSCTL_ADD_INT(ctx, children, OID_AUTO, "ddp_thres", CTLFLAG_RW, &sc->tt.ddp_thres, 0, "DDP threshold"); sc->tt.rx_coalesce = 1; SYSCTL_ADD_INT(ctx, children, OID_AUTO, "rx_coalesce", CTLFLAG_RW, &sc->tt.rx_coalesce, 0, "receive coalescing"); sc->tt.tx_align = 1; SYSCTL_ADD_INT(ctx, children, OID_AUTO, "tx_align", CTLFLAG_RW, &sc->tt.tx_align, 0, "chop and align payload"); } #endif return (0); } static int cxgbe_sysctls(struct port_info *pi) { struct sysctl_ctx_list *ctx; struct sysctl_oid *oid; struct sysctl_oid_list *children; struct adapter *sc = pi->adapter; ctx = device_get_sysctl_ctx(pi->dev); /* * dev.cxgbe.X. */ oid = device_get_sysctl_tree(pi->dev); children = SYSCTL_CHILDREN(oid); SYSCTL_ADD_PROC(ctx, children, OID_AUTO, "linkdnrc", CTLTYPE_STRING | CTLFLAG_RD, pi, 0, sysctl_linkdnrc, "A", "reason why link is down"); if (pi->port_type == FW_PORT_TYPE_BT_XAUI) { SYSCTL_ADD_PROC(ctx, children, OID_AUTO, "temperature", CTLTYPE_INT | CTLFLAG_RD, pi, 0, sysctl_btphy, "I", "PHY temperature (in Celsius)"); SYSCTL_ADD_PROC(ctx, children, OID_AUTO, "fw_version", CTLTYPE_INT | CTLFLAG_RD, pi, 1, sysctl_btphy, "I", "PHY firmware version"); } SYSCTL_ADD_INT(ctx, children, OID_AUTO, "nrxq", CTLFLAG_RD, &pi->nrxq, 0, "# of rx queues"); SYSCTL_ADD_INT(ctx, children, OID_AUTO, "ntxq", CTLFLAG_RD, &pi->ntxq, 0, "# of tx queues"); SYSCTL_ADD_INT(ctx, children, OID_AUTO, "first_rxq", CTLFLAG_RD, &pi->first_rxq, 0, "index of first rx queue"); SYSCTL_ADD_INT(ctx, children, OID_AUTO, "first_txq", CTLFLAG_RD, &pi->first_txq, 0, "index of first tx queue"); SYSCTL_ADD_PROC(ctx, children, OID_AUTO, "rsrv_noflowq", CTLTYPE_INT | CTLFLAG_RW, pi, 0, sysctl_noflowq, "IU", "Reserve queue 0 for non-flowid packets"); #ifdef TCP_OFFLOAD if (is_offload(sc)) { SYSCTL_ADD_INT(ctx, children, OID_AUTO, "nofldrxq", CTLFLAG_RD, &pi->nofldrxq, 0, "# of rx queues for offloaded TCP connections"); SYSCTL_ADD_INT(ctx, children, OID_AUTO, "nofldtxq", CTLFLAG_RD, &pi->nofldtxq, 0, "# of tx queues for offloaded TCP connections"); SYSCTL_ADD_INT(ctx, children, OID_AUTO, "first_ofld_rxq", CTLFLAG_RD, &pi->first_ofld_rxq, 0, "index of first TOE rx queue"); SYSCTL_ADD_INT(ctx, children, OID_AUTO, "first_ofld_txq", CTLFLAG_RD, &pi->first_ofld_txq, 0, "index of first TOE tx queue"); } #endif #ifdef DEV_NETMAP SYSCTL_ADD_INT(ctx, children, OID_AUTO, "nnmrxq", CTLFLAG_RD, &pi->nnmrxq, 0, "# of rx queues for netmap"); SYSCTL_ADD_INT(ctx, children, OID_AUTO, "nnmtxq", CTLFLAG_RD, &pi->nnmtxq, 0, "# of tx queues for netmap"); SYSCTL_ADD_INT(ctx, children, OID_AUTO, "first_nm_rxq", CTLFLAG_RD, &pi->first_nm_rxq, 0, "index of first netmap rx queue"); SYSCTL_ADD_INT(ctx, children, OID_AUTO, "first_nm_txq", CTLFLAG_RD, &pi->first_nm_txq, 0, "index of first netmap tx queue"); #endif SYSCTL_ADD_PROC(ctx, children, OID_AUTO, "holdoff_tmr_idx", CTLTYPE_INT | CTLFLAG_RW, pi, 0, sysctl_holdoff_tmr_idx, "I", "holdoff timer index"); SYSCTL_ADD_PROC(ctx, children, OID_AUTO, "holdoff_pktc_idx", CTLTYPE_INT | CTLFLAG_RW, pi, 0, sysctl_holdoff_pktc_idx, "I", "holdoff packet counter index"); SYSCTL_ADD_PROC(ctx, children, OID_AUTO, "qsize_rxq", CTLTYPE_INT | CTLFLAG_RW, pi, 0, sysctl_qsize_rxq, "I", "rx queue size"); SYSCTL_ADD_PROC(ctx, children, OID_AUTO, "qsize_txq", CTLTYPE_INT | CTLFLAG_RW, pi, 0, sysctl_qsize_txq, "I", "tx queue size"); SYSCTL_ADD_PROC(ctx, children, OID_AUTO, "pause_settings", CTLTYPE_STRING | CTLFLAG_RW, pi, PAUSE_TX, sysctl_pause_settings, "A", "PAUSE settings (bit 0 = rx_pause, bit 1 = tx_pause)"); /* * dev.cxgbe.X.stats. */ oid = SYSCTL_ADD_NODE(ctx, children, OID_AUTO, "stats", CTLFLAG_RD, NULL, "port statistics"); children = SYSCTL_CHILDREN(oid); SYSCTL_ADD_UINT(ctx, children, OID_AUTO, "tx_parse_error", CTLFLAG_RD, &pi->tx_parse_error, 0, "# of tx packets with invalid length or # of segments"); #define SYSCTL_ADD_T4_REG64(pi, name, desc, reg) \ SYSCTL_ADD_OID(ctx, children, OID_AUTO, name, \ CTLTYPE_U64 | CTLFLAG_RD, sc, reg, \ sysctl_handle_t4_reg64, "QU", desc) SYSCTL_ADD_T4_REG64(pi, "tx_octets", "# of octets in good frames", PORT_REG(pi->tx_chan, A_MPS_PORT_STAT_TX_PORT_BYTES_L)); SYSCTL_ADD_T4_REG64(pi, "tx_frames", "total # of good frames", PORT_REG(pi->tx_chan, A_MPS_PORT_STAT_TX_PORT_FRAMES_L)); SYSCTL_ADD_T4_REG64(pi, "tx_bcast_frames", "# of broadcast frames", PORT_REG(pi->tx_chan, A_MPS_PORT_STAT_TX_PORT_BCAST_L)); SYSCTL_ADD_T4_REG64(pi, "tx_mcast_frames", "# of multicast frames", PORT_REG(pi->tx_chan, A_MPS_PORT_STAT_TX_PORT_MCAST_L)); SYSCTL_ADD_T4_REG64(pi, "tx_ucast_frames", "# of unicast frames", PORT_REG(pi->tx_chan, A_MPS_PORT_STAT_TX_PORT_UCAST_L)); SYSCTL_ADD_T4_REG64(pi, "tx_error_frames", "# of error frames", PORT_REG(pi->tx_chan, A_MPS_PORT_STAT_TX_PORT_ERROR_L)); SYSCTL_ADD_T4_REG64(pi, "tx_frames_64", "# of tx frames in this range", PORT_REG(pi->tx_chan, A_MPS_PORT_STAT_TX_PORT_64B_L)); SYSCTL_ADD_T4_REG64(pi, "tx_frames_65_127", "# of tx frames in this range", PORT_REG(pi->tx_chan, A_MPS_PORT_STAT_TX_PORT_65B_127B_L)); SYSCTL_ADD_T4_REG64(pi, "tx_frames_128_255", "# of tx frames in this range", PORT_REG(pi->tx_chan, A_MPS_PORT_STAT_TX_PORT_128B_255B_L)); SYSCTL_ADD_T4_REG64(pi, "tx_frames_256_511", "# of tx frames in this range", PORT_REG(pi->tx_chan, A_MPS_PORT_STAT_TX_PORT_256B_511B_L)); SYSCTL_ADD_T4_REG64(pi, "tx_frames_512_1023", "# of tx frames in this range", PORT_REG(pi->tx_chan, A_MPS_PORT_STAT_TX_PORT_512B_1023B_L)); SYSCTL_ADD_T4_REG64(pi, "tx_frames_1024_1518", "# of tx frames in this range", PORT_REG(pi->tx_chan, A_MPS_PORT_STAT_TX_PORT_1024B_1518B_L)); SYSCTL_ADD_T4_REG64(pi, "tx_frames_1519_max", "# of tx frames in this range", PORT_REG(pi->tx_chan, A_MPS_PORT_STAT_TX_PORT_1519B_MAX_L)); SYSCTL_ADD_T4_REG64(pi, "tx_drop", "# of dropped tx frames", PORT_REG(pi->tx_chan, A_MPS_PORT_STAT_TX_PORT_DROP_L)); SYSCTL_ADD_T4_REG64(pi, "tx_pause", "# of pause frames transmitted", PORT_REG(pi->tx_chan, A_MPS_PORT_STAT_TX_PORT_PAUSE_L)); SYSCTL_ADD_T4_REG64(pi, "tx_ppp0", "# of PPP prio 0 frames transmitted", PORT_REG(pi->tx_chan, A_MPS_PORT_STAT_TX_PORT_PPP0_L)); SYSCTL_ADD_T4_REG64(pi, "tx_ppp1", "# of PPP prio 1 frames transmitted", PORT_REG(pi->tx_chan, A_MPS_PORT_STAT_TX_PORT_PPP1_L)); SYSCTL_ADD_T4_REG64(pi, "tx_ppp2", "# of PPP prio 2 frames transmitted", PORT_REG(pi->tx_chan, A_MPS_PORT_STAT_TX_PORT_PPP2_L)); SYSCTL_ADD_T4_REG64(pi, "tx_ppp3", "# of PPP prio 3 frames transmitted", PORT_REG(pi->tx_chan, A_MPS_PORT_STAT_TX_PORT_PPP3_L)); SYSCTL_ADD_T4_REG64(pi, "tx_ppp4", "# of PPP prio 4 frames transmitted", PORT_REG(pi->tx_chan, A_MPS_PORT_STAT_TX_PORT_PPP4_L)); SYSCTL_ADD_T4_REG64(pi, "tx_ppp5", "# of PPP prio 5 frames transmitted", PORT_REG(pi->tx_chan, A_MPS_PORT_STAT_TX_PORT_PPP5_L)); SYSCTL_ADD_T4_REG64(pi, "tx_ppp6", "# of PPP prio 6 frames transmitted", PORT_REG(pi->tx_chan, A_MPS_PORT_STAT_TX_PORT_PPP6_L)); SYSCTL_ADD_T4_REG64(pi, "tx_ppp7", "# of PPP prio 7 frames transmitted", PORT_REG(pi->tx_chan, A_MPS_PORT_STAT_TX_PORT_PPP7_L)); SYSCTL_ADD_T4_REG64(pi, "rx_octets", "# of octets in good frames", PORT_REG(pi->tx_chan, A_MPS_PORT_STAT_RX_PORT_BYTES_L)); SYSCTL_ADD_T4_REG64(pi, "rx_frames", "total # of good frames", PORT_REG(pi->tx_chan, A_MPS_PORT_STAT_RX_PORT_FRAMES_L)); SYSCTL_ADD_T4_REG64(pi, "rx_bcast_frames", "# of broadcast frames", PORT_REG(pi->tx_chan, A_MPS_PORT_STAT_RX_PORT_BCAST_L)); SYSCTL_ADD_T4_REG64(pi, "rx_mcast_frames", "# of multicast frames", PORT_REG(pi->tx_chan, A_MPS_PORT_STAT_RX_PORT_MCAST_L)); SYSCTL_ADD_T4_REG64(pi, "rx_ucast_frames", "# of unicast frames", PORT_REG(pi->tx_chan, A_MPS_PORT_STAT_RX_PORT_UCAST_L)); SYSCTL_ADD_T4_REG64(pi, "rx_too_long", "# of frames exceeding MTU", PORT_REG(pi->tx_chan, A_MPS_PORT_STAT_RX_PORT_MTU_ERROR_L)); SYSCTL_ADD_T4_REG64(pi, "rx_jabber", "# of jabber frames", PORT_REG(pi->tx_chan, A_MPS_PORT_STAT_RX_PORT_MTU_CRC_ERROR_L)); SYSCTL_ADD_T4_REG64(pi, "rx_fcs_err", "# of frames received with bad FCS", PORT_REG(pi->tx_chan, A_MPS_PORT_STAT_RX_PORT_CRC_ERROR_L)); SYSCTL_ADD_T4_REG64(pi, "rx_len_err", "# of frames received with length error", PORT_REG(pi->tx_chan, A_MPS_PORT_STAT_RX_PORT_LEN_ERROR_L)); SYSCTL_ADD_T4_REG64(pi, "rx_symbol_err", "symbol errors", PORT_REG(pi->tx_chan, A_MPS_PORT_STAT_RX_PORT_SYM_ERROR_L)); SYSCTL_ADD_T4_REG64(pi, "rx_runt", "# of short frames received", PORT_REG(pi->tx_chan, A_MPS_PORT_STAT_RX_PORT_LESS_64B_L)); SYSCTL_ADD_T4_REG64(pi, "rx_frames_64", "# of rx frames in this range", PORT_REG(pi->tx_chan, A_MPS_PORT_STAT_RX_PORT_64B_L)); SYSCTL_ADD_T4_REG64(pi, "rx_frames_65_127", "# of rx frames in this range", PORT_REG(pi->tx_chan, A_MPS_PORT_STAT_RX_PORT_65B_127B_L)); SYSCTL_ADD_T4_REG64(pi, "rx_frames_128_255", "# of rx frames in this range", PORT_REG(pi->tx_chan, A_MPS_PORT_STAT_RX_PORT_128B_255B_L)); SYSCTL_ADD_T4_REG64(pi, "rx_frames_256_511", "# of rx frames in this range", PORT_REG(pi->tx_chan, A_MPS_PORT_STAT_RX_PORT_256B_511B_L)); SYSCTL_ADD_T4_REG64(pi, "rx_frames_512_1023", "# of rx frames in this range", PORT_REG(pi->tx_chan, A_MPS_PORT_STAT_RX_PORT_512B_1023B_L)); SYSCTL_ADD_T4_REG64(pi, "rx_frames_1024_1518", "# of rx frames in this range", PORT_REG(pi->tx_chan, A_MPS_PORT_STAT_RX_PORT_1024B_1518B_L)); SYSCTL_ADD_T4_REG64(pi, "rx_frames_1519_max", "# of rx frames in this range", PORT_REG(pi->tx_chan, A_MPS_PORT_STAT_RX_PORT_1519B_MAX_L)); SYSCTL_ADD_T4_REG64(pi, "rx_pause", "# of pause frames received", PORT_REG(pi->tx_chan, A_MPS_PORT_STAT_RX_PORT_PAUSE_L)); SYSCTL_ADD_T4_REG64(pi, "rx_ppp0", "# of PPP prio 0 frames received", PORT_REG(pi->tx_chan, A_MPS_PORT_STAT_RX_PORT_PPP0_L)); SYSCTL_ADD_T4_REG64(pi, "rx_ppp1", "# of PPP prio 1 frames received", PORT_REG(pi->tx_chan, A_MPS_PORT_STAT_RX_PORT_PPP1_L)); SYSCTL_ADD_T4_REG64(pi, "rx_ppp2", "# of PPP prio 2 frames received", PORT_REG(pi->tx_chan, A_MPS_PORT_STAT_RX_PORT_PPP2_L)); SYSCTL_ADD_T4_REG64(pi, "rx_ppp3", "# of PPP prio 3 frames received", PORT_REG(pi->tx_chan, A_MPS_PORT_STAT_RX_PORT_PPP3_L)); SYSCTL_ADD_T4_REG64(pi, "rx_ppp4", "# of PPP prio 4 frames received", PORT_REG(pi->tx_chan, A_MPS_PORT_STAT_RX_PORT_PPP4_L)); SYSCTL_ADD_T4_REG64(pi, "rx_ppp5", "# of PPP prio 5 frames received", PORT_REG(pi->tx_chan, A_MPS_PORT_STAT_RX_PORT_PPP5_L)); SYSCTL_ADD_T4_REG64(pi, "rx_ppp6", "# of PPP prio 6 frames received", PORT_REG(pi->tx_chan, A_MPS_PORT_STAT_RX_PORT_PPP6_L)); SYSCTL_ADD_T4_REG64(pi, "rx_ppp7", "# of PPP prio 7 frames received", PORT_REG(pi->tx_chan, A_MPS_PORT_STAT_RX_PORT_PPP7_L)); #undef SYSCTL_ADD_T4_REG64 #define SYSCTL_ADD_T4_PORTSTAT(name, desc) \ SYSCTL_ADD_UQUAD(ctx, children, OID_AUTO, #name, CTLFLAG_RD, \ &pi->stats.name, desc) /* We get these from port_stats and they may be stale by upto 1s */ SYSCTL_ADD_T4_PORTSTAT(rx_ovflow0, "# drops due to buffer-group 0 overflows"); SYSCTL_ADD_T4_PORTSTAT(rx_ovflow1, "# drops due to buffer-group 1 overflows"); SYSCTL_ADD_T4_PORTSTAT(rx_ovflow2, "# drops due to buffer-group 2 overflows"); SYSCTL_ADD_T4_PORTSTAT(rx_ovflow3, "# drops due to buffer-group 3 overflows"); SYSCTL_ADD_T4_PORTSTAT(rx_trunc0, "# of buffer-group 0 truncated packets"); SYSCTL_ADD_T4_PORTSTAT(rx_trunc1, "# of buffer-group 1 truncated packets"); SYSCTL_ADD_T4_PORTSTAT(rx_trunc2, "# of buffer-group 2 truncated packets"); SYSCTL_ADD_T4_PORTSTAT(rx_trunc3, "# of buffer-group 3 truncated packets"); #undef SYSCTL_ADD_T4_PORTSTAT return (0); } static int sysctl_int_array(SYSCTL_HANDLER_ARGS) { int rc, *i, space = 0; struct sbuf sb; sbuf_new_for_sysctl(&sb, NULL, 64, req); for (i = arg1; arg2; arg2 -= sizeof(int), i++) { if (space) sbuf_printf(&sb, " "); sbuf_printf(&sb, "%d", *i); space = 1; } rc = sbuf_finish(&sb); sbuf_delete(&sb); return (rc); } static int sysctl_bitfield(SYSCTL_HANDLER_ARGS) { int rc; struct sbuf *sb; rc = sysctl_wire_old_buffer(req, 0); if (rc != 0) return(rc); sb = sbuf_new_for_sysctl(NULL, NULL, 128, req); if (sb == NULL) return (ENOMEM); sbuf_printf(sb, "%b", (int)arg2, (char *)arg1); rc = sbuf_finish(sb); sbuf_delete(sb); return (rc); } static int sysctl_btphy(SYSCTL_HANDLER_ARGS) { struct port_info *pi = arg1; int op = arg2; struct adapter *sc = pi->adapter; u_int v; int rc; rc = begin_synchronized_op(sc, pi, SLEEP_OK | INTR_OK, "t4btt"); if (rc) return (rc); /* XXX: magic numbers */ rc = -t4_mdio_rd(sc, sc->mbox, pi->mdio_addr, 0x1e, op ? 0x20 : 0xc820, &v); end_synchronized_op(sc, 0); if (rc) return (rc); if (op == 0) v /= 256; rc = sysctl_handle_int(oidp, &v, 0, req); return (rc); } static int sysctl_noflowq(SYSCTL_HANDLER_ARGS) { struct port_info *pi = arg1; int rc, val; val = pi->rsrv_noflowq; rc = sysctl_handle_int(oidp, &val, 0, req); if (rc != 0 || req->newptr == NULL) return (rc); if ((val >= 1) && (pi->ntxq > 1)) pi->rsrv_noflowq = 1; else pi->rsrv_noflowq = 0; return (rc); } static int sysctl_holdoff_tmr_idx(SYSCTL_HANDLER_ARGS) { struct port_info *pi = arg1; struct adapter *sc = pi->adapter; int idx, rc, i; struct sge_rxq *rxq; #ifdef TCP_OFFLOAD struct sge_ofld_rxq *ofld_rxq; #endif uint8_t v; idx = pi->tmr_idx; rc = sysctl_handle_int(oidp, &idx, 0, req); if (rc != 0 || req->newptr == NULL) return (rc); if (idx < 0 || idx >= SGE_NTIMERS) return (EINVAL); rc = begin_synchronized_op(sc, pi, HOLD_LOCK | SLEEP_OK | INTR_OK, "t4tmr"); if (rc) return (rc); v = V_QINTR_TIMER_IDX(idx) | V_QINTR_CNT_EN(pi->pktc_idx != -1); for_each_rxq(pi, i, rxq) { #ifdef atomic_store_rel_8 atomic_store_rel_8(&rxq->iq.intr_params, v); #else rxq->iq.intr_params = v; #endif } #ifdef TCP_OFFLOAD for_each_ofld_rxq(pi, i, ofld_rxq) { #ifdef atomic_store_rel_8 atomic_store_rel_8(&ofld_rxq->iq.intr_params, v); #else ofld_rxq->iq.intr_params = v; #endif } #endif pi->tmr_idx = idx; end_synchronized_op(sc, LOCK_HELD); return (0); } static int sysctl_holdoff_pktc_idx(SYSCTL_HANDLER_ARGS) { struct port_info *pi = arg1; struct adapter *sc = pi->adapter; int idx, rc; idx = pi->pktc_idx; rc = sysctl_handle_int(oidp, &idx, 0, req); if (rc != 0 || req->newptr == NULL) return (rc); if (idx < -1 || idx >= SGE_NCOUNTERS) return (EINVAL); rc = begin_synchronized_op(sc, pi, HOLD_LOCK | SLEEP_OK | INTR_OK, "t4pktc"); if (rc) return (rc); if (pi->flags & PORT_INIT_DONE) rc = EBUSY; /* cannot be changed once the queues are created */ else pi->pktc_idx = idx; end_synchronized_op(sc, LOCK_HELD); return (rc); } static int sysctl_qsize_rxq(SYSCTL_HANDLER_ARGS) { struct port_info *pi = arg1; struct adapter *sc = pi->adapter; int qsize, rc; qsize = pi->qsize_rxq; rc = sysctl_handle_int(oidp, &qsize, 0, req); if (rc != 0 || req->newptr == NULL) return (rc); if (qsize < 128 || (qsize & 7)) return (EINVAL); rc = begin_synchronized_op(sc, pi, HOLD_LOCK | SLEEP_OK | INTR_OK, "t4rxqs"); if (rc) return (rc); if (pi->flags & PORT_INIT_DONE) rc = EBUSY; /* cannot be changed once the queues are created */ else pi->qsize_rxq = qsize; end_synchronized_op(sc, LOCK_HELD); return (rc); } static int sysctl_qsize_txq(SYSCTL_HANDLER_ARGS) { struct port_info *pi = arg1; struct adapter *sc = pi->adapter; int qsize, rc; qsize = pi->qsize_txq; rc = sysctl_handle_int(oidp, &qsize, 0, req); if (rc != 0 || req->newptr == NULL) return (rc); if (qsize < 128 || qsize > 65536) return (EINVAL); rc = begin_synchronized_op(sc, pi, HOLD_LOCK | SLEEP_OK | INTR_OK, "t4txqs"); if (rc) return (rc); if (pi->flags & PORT_INIT_DONE) rc = EBUSY; /* cannot be changed once the queues are created */ else pi->qsize_txq = qsize; end_synchronized_op(sc, LOCK_HELD); return (rc); } static int sysctl_pause_settings(SYSCTL_HANDLER_ARGS) { struct port_info *pi = arg1; struct adapter *sc = pi->adapter; struct link_config *lc = &pi->link_cfg; int rc; if (req->newptr == NULL) { struct sbuf *sb; static char *bits = "\20\1PAUSE_RX\2PAUSE_TX"; rc = sysctl_wire_old_buffer(req, 0); if (rc != 0) return(rc); sb = sbuf_new_for_sysctl(NULL, NULL, 128, req); if (sb == NULL) return (ENOMEM); sbuf_printf(sb, "%b", lc->fc & (PAUSE_TX | PAUSE_RX), bits); rc = sbuf_finish(sb); sbuf_delete(sb); } else { char s[2]; int n; s[0] = '0' + (lc->requested_fc & (PAUSE_TX | PAUSE_RX)); s[1] = 0; rc = sysctl_handle_string(oidp, s, sizeof(s), req); if (rc != 0) return(rc); if (s[1] != 0) return (EINVAL); if (s[0] < '0' || s[0] > '9') return (EINVAL); /* not a number */ n = s[0] - '0'; if (n & ~(PAUSE_TX | PAUSE_RX)) return (EINVAL); /* some other bit is set too */ rc = begin_synchronized_op(sc, pi, SLEEP_OK | INTR_OK, "t4PAUSE"); if (rc) return (rc); if ((lc->requested_fc & (PAUSE_TX | PAUSE_RX)) != n) { int link_ok = lc->link_ok; lc->requested_fc &= ~(PAUSE_TX | PAUSE_RX); lc->requested_fc |= n; rc = -t4_link_start(sc, sc->mbox, pi->tx_chan, lc); lc->link_ok = link_ok; /* restore */ } end_synchronized_op(sc, 0); } return (rc); } static int sysctl_handle_t4_reg64(SYSCTL_HANDLER_ARGS) { struct adapter *sc = arg1; int reg = arg2; uint64_t val; val = t4_read_reg64(sc, reg); return (sysctl_handle_64(oidp, &val, 0, req)); } static int sysctl_temperature(SYSCTL_HANDLER_ARGS) { struct adapter *sc = arg1; int rc, t; uint32_t param, val; rc = begin_synchronized_op(sc, NULL, SLEEP_OK | INTR_OK, "t4temp"); if (rc) return (rc); param = V_FW_PARAMS_MNEM(FW_PARAMS_MNEM_DEV) | V_FW_PARAMS_PARAM_X(FW_PARAMS_PARAM_DEV_DIAG) | V_FW_PARAMS_PARAM_Y(FW_PARAM_DEV_DIAG_TMP); rc = -t4_query_params(sc, sc->mbox, sc->pf, 0, 1, ¶m, &val); end_synchronized_op(sc, 0); if (rc) return (rc); /* unknown is returned as 0 but we display -1 in that case */ t = val == 0 ? -1 : val; rc = sysctl_handle_int(oidp, &t, 0, req); return (rc); } #ifdef SBUF_DRAIN static int sysctl_cctrl(SYSCTL_HANDLER_ARGS) { struct adapter *sc = arg1; struct sbuf *sb; int rc, i; uint16_t incr[NMTUS][NCCTRL_WIN]; static const char *dec_fac[] = { "0.5", "0.5625", "0.625", "0.6875", "0.75", "0.8125", "0.875", "0.9375" }; rc = sysctl_wire_old_buffer(req, 0); if (rc != 0) return (rc); sb = sbuf_new_for_sysctl(NULL, NULL, 4096, req); if (sb == NULL) return (ENOMEM); t4_read_cong_tbl(sc, incr); for (i = 0; i < NCCTRL_WIN; ++i) { sbuf_printf(sb, "%2d: %4u %4u %4u %4u %4u %4u %4u %4u\n", i, incr[0][i], incr[1][i], incr[2][i], incr[3][i], incr[4][i], incr[5][i], incr[6][i], incr[7][i]); sbuf_printf(sb, "%8u %4u %4u %4u %4u %4u %4u %4u %5u %s\n", incr[8][i], incr[9][i], incr[10][i], incr[11][i], incr[12][i], incr[13][i], incr[14][i], incr[15][i], sc->params.a_wnd[i], dec_fac[sc->params.b_wnd[i]]); } rc = sbuf_finish(sb); sbuf_delete(sb); return (rc); } static const char *qname[CIM_NUM_IBQ + CIM_NUM_OBQ_T5] = { "TP0", "TP1", "ULP", "SGE0", "SGE1", "NC-SI", /* ibq's */ "ULP0", "ULP1", "ULP2", "ULP3", "SGE", "NC-SI", /* obq's */ "SGE0-RX", "SGE1-RX" /* additional obq's (T5 onwards) */ }; static int sysctl_cim_ibq_obq(SYSCTL_HANDLER_ARGS) { struct adapter *sc = arg1; struct sbuf *sb; int rc, i, n, qid = arg2; uint32_t *buf, *p; char *qtype; u_int cim_num_obq = is_t4(sc) ? CIM_NUM_OBQ : CIM_NUM_OBQ_T5; KASSERT(qid >= 0 && qid < CIM_NUM_IBQ + cim_num_obq, ("%s: bad qid %d\n", __func__, qid)); if (qid < CIM_NUM_IBQ) { /* inbound queue */ qtype = "IBQ"; n = 4 * CIM_IBQ_SIZE; buf = malloc(n * sizeof(uint32_t), M_CXGBE, M_ZERO | M_WAITOK); rc = t4_read_cim_ibq(sc, qid, buf, n); } else { /* outbound queue */ qtype = "OBQ"; qid -= CIM_NUM_IBQ; n = 4 * cim_num_obq * CIM_OBQ_SIZE; buf = malloc(n * sizeof(uint32_t), M_CXGBE, M_ZERO | M_WAITOK); rc = t4_read_cim_obq(sc, qid, buf, n); } if (rc < 0) { rc = -rc; goto done; } n = rc * sizeof(uint32_t); /* rc has # of words actually read */ rc = sysctl_wire_old_buffer(req, 0); if (rc != 0) goto done; sb = sbuf_new_for_sysctl(NULL, NULL, PAGE_SIZE, req); if (sb == NULL) { rc = ENOMEM; goto done; } sbuf_printf(sb, "%s%d %s", qtype , qid, qname[arg2]); for (i = 0, p = buf; i < n; i += 16, p += 4) sbuf_printf(sb, "\n%#06x: %08x %08x %08x %08x", i, p[0], p[1], p[2], p[3]); rc = sbuf_finish(sb); sbuf_delete(sb); done: free(buf, M_CXGBE); return (rc); } static int sysctl_cim_la(SYSCTL_HANDLER_ARGS) { struct adapter *sc = arg1; u_int cfg; struct sbuf *sb; uint32_t *buf, *p; int rc; rc = -t4_cim_read(sc, A_UP_UP_DBG_LA_CFG, 1, &cfg); if (rc != 0) return (rc); rc = sysctl_wire_old_buffer(req, 0); if (rc != 0) return (rc); sb = sbuf_new_for_sysctl(NULL, NULL, 4096, req); if (sb == NULL) return (ENOMEM); buf = malloc(sc->params.cim_la_size * sizeof(uint32_t), M_CXGBE, M_ZERO | M_WAITOK); rc = -t4_cim_read_la(sc, buf, NULL); if (rc != 0) goto done; sbuf_printf(sb, "Status Data PC%s", cfg & F_UPDBGLACAPTPCONLY ? "" : " LS0Stat LS0Addr LS0Data"); KASSERT((sc->params.cim_la_size & 7) == 0, ("%s: p will walk off the end of buf", __func__)); for (p = buf; p < &buf[sc->params.cim_la_size]; p += 8) { if (cfg & F_UPDBGLACAPTPCONLY) { sbuf_printf(sb, "\n %02x %08x %08x", p[5] & 0xff, p[6], p[7]); sbuf_printf(sb, "\n %02x %02x%06x %02x%06x", (p[3] >> 8) & 0xff, p[3] & 0xff, p[4] >> 8, p[4] & 0xff, p[5] >> 8); sbuf_printf(sb, "\n %02x %x%07x %x%07x", (p[0] >> 4) & 0xff, p[0] & 0xf, p[1] >> 4, p[1] & 0xf, p[2] >> 4); } else { sbuf_printf(sb, "\n %02x %x%07x %x%07x %08x %08x " "%08x%08x%08x%08x", (p[0] >> 4) & 0xff, p[0] & 0xf, p[1] >> 4, p[1] & 0xf, p[2] >> 4, p[2] & 0xf, p[3], p[4], p[5], p[6], p[7]); } } rc = sbuf_finish(sb); sbuf_delete(sb); done: free(buf, M_CXGBE); return (rc); } static int sysctl_cim_ma_la(SYSCTL_HANDLER_ARGS) { struct adapter *sc = arg1; u_int i; struct sbuf *sb; uint32_t *buf, *p; int rc; rc = sysctl_wire_old_buffer(req, 0); if (rc != 0) return (rc); sb = sbuf_new_for_sysctl(NULL, NULL, 4096, req); if (sb == NULL) return (ENOMEM); buf = malloc(2 * CIM_MALA_SIZE * 5 * sizeof(uint32_t), M_CXGBE, M_ZERO | M_WAITOK); t4_cim_read_ma_la(sc, buf, buf + 5 * CIM_MALA_SIZE); p = buf; for (i = 0; i < CIM_MALA_SIZE; i++, p += 5) { sbuf_printf(sb, "\n%02x%08x%08x%08x%08x", p[4], p[3], p[2], p[1], p[0]); } sbuf_printf(sb, "\n\nCnt ID Tag UE Data RDY VLD"); for (i = 0; i < CIM_MALA_SIZE; i++, p += 5) { sbuf_printf(sb, "\n%3u %2u %x %u %08x%08x %u %u", (p[2] >> 10) & 0xff, (p[2] >> 7) & 7, (p[2] >> 3) & 0xf, (p[2] >> 2) & 1, (p[1] >> 2) | ((p[2] & 3) << 30), (p[0] >> 2) | ((p[1] & 3) << 30), (p[0] >> 1) & 1, p[0] & 1); } rc = sbuf_finish(sb); sbuf_delete(sb); free(buf, M_CXGBE); return (rc); } static int sysctl_cim_pif_la(SYSCTL_HANDLER_ARGS) { struct adapter *sc = arg1; u_int i; struct sbuf *sb; uint32_t *buf, *p; int rc; rc = sysctl_wire_old_buffer(req, 0); if (rc != 0) return (rc); sb = sbuf_new_for_sysctl(NULL, NULL, 4096, req); if (sb == NULL) return (ENOMEM); buf = malloc(2 * CIM_PIFLA_SIZE * 6 * sizeof(uint32_t), M_CXGBE, M_ZERO | M_WAITOK); t4_cim_read_pif_la(sc, buf, buf + 6 * CIM_PIFLA_SIZE, NULL, NULL); p = buf; sbuf_printf(sb, "Cntl ID DataBE Addr Data"); for (i = 0; i < CIM_MALA_SIZE; i++, p += 6) { sbuf_printf(sb, "\n %02x %02x %04x %08x %08x%08x%08x%08x", (p[5] >> 22) & 0xff, (p[5] >> 16) & 0x3f, p[5] & 0xffff, p[4], p[3], p[2], p[1], p[0]); } sbuf_printf(sb, "\n\nCntl ID Data"); for (i = 0; i < CIM_MALA_SIZE; i++, p += 6) { sbuf_printf(sb, "\n %02x %02x %08x%08x%08x%08x", (p[4] >> 6) & 0xff, p[4] & 0x3f, p[3], p[2], p[1], p[0]); } rc = sbuf_finish(sb); sbuf_delete(sb); free(buf, M_CXGBE); return (rc); } static int sysctl_cim_qcfg(SYSCTL_HANDLER_ARGS) { struct adapter *sc = arg1; struct sbuf *sb; int rc, i; uint16_t base[CIM_NUM_IBQ + CIM_NUM_OBQ_T5]; uint16_t size[CIM_NUM_IBQ + CIM_NUM_OBQ_T5]; uint16_t thres[CIM_NUM_IBQ]; uint32_t obq_wr[2 * CIM_NUM_OBQ_T5], *wr = obq_wr; uint32_t stat[4 * (CIM_NUM_IBQ + CIM_NUM_OBQ_T5)], *p = stat; u_int cim_num_obq, ibq_rdaddr, obq_rdaddr, nq; if (is_t4(sc)) { cim_num_obq = CIM_NUM_OBQ; ibq_rdaddr = A_UP_IBQ_0_RDADDR; obq_rdaddr = A_UP_OBQ_0_REALADDR; } else { cim_num_obq = CIM_NUM_OBQ_T5; ibq_rdaddr = A_UP_IBQ_0_SHADOW_RDADDR; obq_rdaddr = A_UP_OBQ_0_SHADOW_REALADDR; } nq = CIM_NUM_IBQ + cim_num_obq; rc = -t4_cim_read(sc, ibq_rdaddr, 4 * nq, stat); if (rc == 0) rc = -t4_cim_read(sc, obq_rdaddr, 2 * cim_num_obq, obq_wr); if (rc != 0) return (rc); t4_read_cimq_cfg(sc, base, size, thres); rc = sysctl_wire_old_buffer(req, 0); if (rc != 0) return (rc); sb = sbuf_new_for_sysctl(NULL, NULL, PAGE_SIZE, req); if (sb == NULL) return (ENOMEM); sbuf_printf(sb, "Queue Base Size Thres RdPtr WrPtr SOP EOP Avail"); for (i = 0; i < CIM_NUM_IBQ; i++, p += 4) sbuf_printf(sb, "\n%7s %5x %5u %5u %6x %4x %4u %4u %5u", qname[i], base[i], size[i], thres[i], G_IBQRDADDR(p[0]), G_IBQWRADDR(p[1]), G_QUESOPCNT(p[3]), G_QUEEOPCNT(p[3]), G_QUEREMFLITS(p[2]) * 16); for ( ; i < nq; i++, p += 4, wr += 2) sbuf_printf(sb, "\n%7s %5x %5u %12x %4x %4u %4u %5u", qname[i], base[i], size[i], G_QUERDADDR(p[0]) & 0x3fff, wr[0] - base[i], G_QUESOPCNT(p[3]), G_QUEEOPCNT(p[3]), G_QUEREMFLITS(p[2]) * 16); rc = sbuf_finish(sb); sbuf_delete(sb); return (rc); } static int sysctl_cpl_stats(SYSCTL_HANDLER_ARGS) { struct adapter *sc = arg1; struct sbuf *sb; int rc; struct tp_cpl_stats stats; rc = sysctl_wire_old_buffer(req, 0); if (rc != 0) return (rc); sb = sbuf_new_for_sysctl(NULL, NULL, 256, req); if (sb == NULL) return (ENOMEM); t4_tp_get_cpl_stats(sc, &stats); sbuf_printf(sb, " channel 0 channel 1 channel 2 " "channel 3\n"); sbuf_printf(sb, "CPL requests: %10u %10u %10u %10u\n", stats.req[0], stats.req[1], stats.req[2], stats.req[3]); sbuf_printf(sb, "CPL responses: %10u %10u %10u %10u", stats.rsp[0], stats.rsp[1], stats.rsp[2], stats.rsp[3]); rc = sbuf_finish(sb); sbuf_delete(sb); return (rc); } static int sysctl_ddp_stats(SYSCTL_HANDLER_ARGS) { struct adapter *sc = arg1; struct sbuf *sb; int rc; struct tp_usm_stats stats; rc = sysctl_wire_old_buffer(req, 0); if (rc != 0) return(rc); sb = sbuf_new_for_sysctl(NULL, NULL, 256, req); if (sb == NULL) return (ENOMEM); t4_get_usm_stats(sc, &stats); sbuf_printf(sb, "Frames: %u\n", stats.frames); sbuf_printf(sb, "Octets: %ju\n", stats.octets); sbuf_printf(sb, "Drops: %u", stats.drops); rc = sbuf_finish(sb); sbuf_delete(sb); return (rc); } const char *devlog_level_strings[] = { [FW_DEVLOG_LEVEL_EMERG] = "EMERG", [FW_DEVLOG_LEVEL_CRIT] = "CRIT", [FW_DEVLOG_LEVEL_ERR] = "ERR", [FW_DEVLOG_LEVEL_NOTICE] = "NOTICE", [FW_DEVLOG_LEVEL_INFO] = "INFO", [FW_DEVLOG_LEVEL_DEBUG] = "DEBUG" }; const char *devlog_facility_strings[] = { [FW_DEVLOG_FACILITY_CORE] = "CORE", [FW_DEVLOG_FACILITY_CF] = "CF", [FW_DEVLOG_FACILITY_SCHED] = "SCHED", [FW_DEVLOG_FACILITY_TIMER] = "TIMER", [FW_DEVLOG_FACILITY_RES] = "RES", [FW_DEVLOG_FACILITY_HW] = "HW", [FW_DEVLOG_FACILITY_FLR] = "FLR", [FW_DEVLOG_FACILITY_DMAQ] = "DMAQ", [FW_DEVLOG_FACILITY_PHY] = "PHY", [FW_DEVLOG_FACILITY_MAC] = "MAC", [FW_DEVLOG_FACILITY_PORT] = "PORT", [FW_DEVLOG_FACILITY_VI] = "VI", [FW_DEVLOG_FACILITY_FILTER] = "FILTER", [FW_DEVLOG_FACILITY_ACL] = "ACL", [FW_DEVLOG_FACILITY_TM] = "TM", [FW_DEVLOG_FACILITY_QFC] = "QFC", [FW_DEVLOG_FACILITY_DCB] = "DCB", [FW_DEVLOG_FACILITY_ETH] = "ETH", [FW_DEVLOG_FACILITY_OFLD] = "OFLD", [FW_DEVLOG_FACILITY_RI] = "RI", [FW_DEVLOG_FACILITY_ISCSI] = "ISCSI", [FW_DEVLOG_FACILITY_FCOE] = "FCOE", [FW_DEVLOG_FACILITY_FOISCSI] = "FOISCSI", [FW_DEVLOG_FACILITY_FOFCOE] = "FOFCOE" }; static int sysctl_devlog(SYSCTL_HANDLER_ARGS) { struct adapter *sc = arg1; struct devlog_params *dparams = &sc->params.devlog; struct fw_devlog_e *buf, *e; int i, j, rc, nentries, first = 0, m; struct sbuf *sb; uint64_t ftstamp = UINT64_MAX; if (dparams->start == 0) { dparams->memtype = FW_MEMTYPE_EDC0; dparams->start = 0x84000; dparams->size = 32768; } nentries = dparams->size / sizeof(struct fw_devlog_e); buf = malloc(dparams->size, M_CXGBE, M_NOWAIT); if (buf == NULL) return (ENOMEM); m = fwmtype_to_hwmtype(dparams->memtype); rc = -t4_mem_read(sc, m, dparams->start, dparams->size, (void *)buf); if (rc != 0) goto done; for (i = 0; i < nentries; i++) { e = &buf[i]; if (e->timestamp == 0) break; /* end */ e->timestamp = be64toh(e->timestamp); e->seqno = be32toh(e->seqno); for (j = 0; j < 8; j++) e->params[j] = be32toh(e->params[j]); if (e->timestamp < ftstamp) { ftstamp = e->timestamp; first = i; } } if (buf[first].timestamp == 0) goto done; /* nothing in the log */ rc = sysctl_wire_old_buffer(req, 0); if (rc != 0) goto done; sb = sbuf_new_for_sysctl(NULL, NULL, 4096, req); if (sb == NULL) { rc = ENOMEM; goto done; } sbuf_printf(sb, "%10s %15s %8s %8s %s\n", "Seq#", "Tstamp", "Level", "Facility", "Message"); i = first; do { e = &buf[i]; if (e->timestamp == 0) break; /* end */ sbuf_printf(sb, "%10d %15ju %8s %8s ", e->seqno, e->timestamp, (e->level < nitems(devlog_level_strings) ? devlog_level_strings[e->level] : "UNKNOWN"), (e->facility < nitems(devlog_facility_strings) ? devlog_facility_strings[e->facility] : "UNKNOWN")); sbuf_printf(sb, e->fmt, e->params[0], e->params[1], e->params[2], e->params[3], e->params[4], e->params[5], e->params[6], e->params[7]); if (++i == nentries) i = 0; } while (i != first); rc = sbuf_finish(sb); sbuf_delete(sb); done: free(buf, M_CXGBE); return (rc); } static int sysctl_fcoe_stats(SYSCTL_HANDLER_ARGS) { struct adapter *sc = arg1; struct sbuf *sb; int rc; struct tp_fcoe_stats stats[4]; rc = sysctl_wire_old_buffer(req, 0); if (rc != 0) return (rc); sb = sbuf_new_for_sysctl(NULL, NULL, 256, req); if (sb == NULL) return (ENOMEM); t4_get_fcoe_stats(sc, 0, &stats[0]); t4_get_fcoe_stats(sc, 1, &stats[1]); t4_get_fcoe_stats(sc, 2, &stats[2]); t4_get_fcoe_stats(sc, 3, &stats[3]); sbuf_printf(sb, " channel 0 channel 1 " "channel 2 channel 3\n"); sbuf_printf(sb, "octetsDDP: %16ju %16ju %16ju %16ju\n", stats[0].octetsDDP, stats[1].octetsDDP, stats[2].octetsDDP, stats[3].octetsDDP); sbuf_printf(sb, "framesDDP: %16u %16u %16u %16u\n", stats[0].framesDDP, stats[1].framesDDP, stats[2].framesDDP, stats[3].framesDDP); sbuf_printf(sb, "framesDrop: %16u %16u %16u %16u", stats[0].framesDrop, stats[1].framesDrop, stats[2].framesDrop, stats[3].framesDrop); rc = sbuf_finish(sb); sbuf_delete(sb); return (rc); } static int sysctl_hw_sched(SYSCTL_HANDLER_ARGS) { struct adapter *sc = arg1; struct sbuf *sb; int rc, i; unsigned int map, kbps, ipg, mode; unsigned int pace_tab[NTX_SCHED]; rc = sysctl_wire_old_buffer(req, 0); if (rc != 0) return (rc); sb = sbuf_new_for_sysctl(NULL, NULL, 256, req); if (sb == NULL) return (ENOMEM); map = t4_read_reg(sc, A_TP_TX_MOD_QUEUE_REQ_MAP); mode = G_TIMERMODE(t4_read_reg(sc, A_TP_MOD_CONFIG)); t4_read_pace_tbl(sc, pace_tab); sbuf_printf(sb, "Scheduler Mode Channel Rate (Kbps) " "Class IPG (0.1 ns) Flow IPG (us)"); for (i = 0; i < NTX_SCHED; ++i, map >>= 2) { t4_get_tx_sched(sc, i, &kbps, &ipg); sbuf_printf(sb, "\n %u %-5s %u ", i, (mode & (1 << i)) ? "flow" : "class", map & 3); if (kbps) sbuf_printf(sb, "%9u ", kbps); else sbuf_printf(sb, " disabled "); if (ipg) sbuf_printf(sb, "%13u ", ipg); else sbuf_printf(sb, " disabled "); if (pace_tab[i]) sbuf_printf(sb, "%10u", pace_tab[i]); else sbuf_printf(sb, " disabled"); } rc = sbuf_finish(sb); sbuf_delete(sb); return (rc); } static int sysctl_lb_stats(SYSCTL_HANDLER_ARGS) { struct adapter *sc = arg1; struct sbuf *sb; int rc, i, j; uint64_t *p0, *p1; struct lb_port_stats s[2]; static const char *stat_name[] = { "OctetsOK:", "FramesOK:", "BcastFrames:", "McastFrames:", "UcastFrames:", "ErrorFrames:", "Frames64:", "Frames65To127:", "Frames128To255:", "Frames256To511:", "Frames512To1023:", "Frames1024To1518:", "Frames1519ToMax:", "FramesDropped:", "BG0FramesDropped:", "BG1FramesDropped:", "BG2FramesDropped:", "BG3FramesDropped:", "BG0FramesTrunc:", "BG1FramesTrunc:", "BG2FramesTrunc:", "BG3FramesTrunc:" }; rc = sysctl_wire_old_buffer(req, 0); if (rc != 0) return (rc); sb = sbuf_new_for_sysctl(NULL, NULL, 4096, req); if (sb == NULL) return (ENOMEM); memset(s, 0, sizeof(s)); for (i = 0; i < 4; i += 2) { t4_get_lb_stats(sc, i, &s[0]); t4_get_lb_stats(sc, i + 1, &s[1]); p0 = &s[0].octets; p1 = &s[1].octets; sbuf_printf(sb, "%s Loopback %u" " Loopback %u", i == 0 ? "" : "\n", i, i + 1); for (j = 0; j < nitems(stat_name); j++) sbuf_printf(sb, "\n%-17s %20ju %20ju", stat_name[j], *p0++, *p1++); } rc = sbuf_finish(sb); sbuf_delete(sb); return (rc); } static int sysctl_linkdnrc(SYSCTL_HANDLER_ARGS) { int rc = 0; struct port_info *pi = arg1; struct sbuf *sb; static const char *linkdnreasons[] = { "non-specific", "remote fault", "autoneg failed", "reserved3", "PHY overheated", "unknown", "rx los", "reserved7" }; rc = sysctl_wire_old_buffer(req, 0); if (rc != 0) return(rc); sb = sbuf_new_for_sysctl(NULL, NULL, 64, req); if (sb == NULL) return (ENOMEM); if (pi->linkdnrc < 0) sbuf_printf(sb, "n/a"); else if (pi->linkdnrc < nitems(linkdnreasons)) sbuf_printf(sb, "%s", linkdnreasons[pi->linkdnrc]); else sbuf_printf(sb, "%d", pi->linkdnrc); rc = sbuf_finish(sb); sbuf_delete(sb); return (rc); } struct mem_desc { unsigned int base; unsigned int limit; unsigned int idx; }; static int mem_desc_cmp(const void *a, const void *b) { return ((const struct mem_desc *)a)->base - ((const struct mem_desc *)b)->base; } static void mem_region_show(struct sbuf *sb, const char *name, unsigned int from, unsigned int to) { unsigned int size; size = to - from + 1; if (size == 0) return; /* XXX: need humanize_number(3) in libkern for a more readable 'size' */ sbuf_printf(sb, "%-15s %#x-%#x [%u]\n", name, from, to, size); } static int sysctl_meminfo(SYSCTL_HANDLER_ARGS) { struct adapter *sc = arg1; struct sbuf *sb; int rc, i, n; uint32_t lo, hi, used, alloc; static const char *memory[] = {"EDC0:", "EDC1:", "MC:", "MC0:", "MC1:"}; static const char *region[] = { "DBQ contexts:", "IMSG contexts:", "FLM cache:", "TCBs:", "Pstructs:", "Timers:", "Rx FL:", "Tx FL:", "Pstruct FL:", "Tx payload:", "Rx payload:", "LE hash:", "iSCSI region:", "TDDP region:", "TPT region:", "STAG region:", "RQ region:", "RQUDP region:", "PBL region:", "TXPBL region:", "DBVFIFO region:", "ULPRX state:", "ULPTX state:", "On-chip queues:" }; struct mem_desc avail[4]; struct mem_desc mem[nitems(region) + 3]; /* up to 3 holes */ struct mem_desc *md = mem; rc = sysctl_wire_old_buffer(req, 0); if (rc != 0) return (rc); sb = sbuf_new_for_sysctl(NULL, NULL, 4096, req); if (sb == NULL) return (ENOMEM); for (i = 0; i < nitems(mem); i++) { mem[i].limit = 0; mem[i].idx = i; } /* Find and sort the populated memory ranges */ i = 0; lo = t4_read_reg(sc, A_MA_TARGET_MEM_ENABLE); if (lo & F_EDRAM0_ENABLE) { hi = t4_read_reg(sc, A_MA_EDRAM0_BAR); avail[i].base = G_EDRAM0_BASE(hi) << 20; avail[i].limit = avail[i].base + (G_EDRAM0_SIZE(hi) << 20); avail[i].idx = 0; i++; } if (lo & F_EDRAM1_ENABLE) { hi = t4_read_reg(sc, A_MA_EDRAM1_BAR); avail[i].base = G_EDRAM1_BASE(hi) << 20; avail[i].limit = avail[i].base + (G_EDRAM1_SIZE(hi) << 20); avail[i].idx = 1; i++; } if (lo & F_EXT_MEM_ENABLE) { hi = t4_read_reg(sc, A_MA_EXT_MEMORY_BAR); avail[i].base = G_EXT_MEM_BASE(hi) << 20; avail[i].limit = avail[i].base + (G_EXT_MEM_SIZE(hi) << 20); avail[i].idx = is_t4(sc) ? 2 : 3; /* Call it MC for T4 */ i++; } if (!is_t4(sc) && lo & F_EXT_MEM1_ENABLE) { hi = t4_read_reg(sc, A_MA_EXT_MEMORY1_BAR); avail[i].base = G_EXT_MEM1_BASE(hi) << 20; avail[i].limit = avail[i].base + (G_EXT_MEM1_SIZE(hi) << 20); avail[i].idx = 4; i++; } if (!i) /* no memory available */ return 0; qsort(avail, i, sizeof(struct mem_desc), mem_desc_cmp); (md++)->base = t4_read_reg(sc, A_SGE_DBQ_CTXT_BADDR); (md++)->base = t4_read_reg(sc, A_SGE_IMSG_CTXT_BADDR); (md++)->base = t4_read_reg(sc, A_SGE_FLM_CACHE_BADDR); (md++)->base = t4_read_reg(sc, A_TP_CMM_TCB_BASE); (md++)->base = t4_read_reg(sc, A_TP_CMM_MM_BASE); (md++)->base = t4_read_reg(sc, A_TP_CMM_TIMER_BASE); (md++)->base = t4_read_reg(sc, A_TP_CMM_MM_RX_FLST_BASE); (md++)->base = t4_read_reg(sc, A_TP_CMM_MM_TX_FLST_BASE); (md++)->base = t4_read_reg(sc, A_TP_CMM_MM_PS_FLST_BASE); /* the next few have explicit upper bounds */ md->base = t4_read_reg(sc, A_TP_PMM_TX_BASE); md->limit = md->base - 1 + t4_read_reg(sc, A_TP_PMM_TX_PAGE_SIZE) * G_PMTXMAXPAGE(t4_read_reg(sc, A_TP_PMM_TX_MAX_PAGE)); md++; md->base = t4_read_reg(sc, A_TP_PMM_RX_BASE); md->limit = md->base - 1 + t4_read_reg(sc, A_TP_PMM_RX_PAGE_SIZE) * G_PMRXMAXPAGE(t4_read_reg(sc, A_TP_PMM_RX_MAX_PAGE)); md++; if (t4_read_reg(sc, A_LE_DB_CONFIG) & F_HASHEN) { hi = t4_read_reg(sc, A_LE_DB_TID_HASHBASE) / 4; md->base = t4_read_reg(sc, A_LE_DB_HASH_TID_BASE); md->limit = (sc->tids.ntids - hi) * 16 + md->base - 1; } else { md->base = 0; md->idx = nitems(region); /* hide it */ } md++; #define ulp_region(reg) \ md->base = t4_read_reg(sc, A_ULP_ ## reg ## _LLIMIT);\ (md++)->limit = t4_read_reg(sc, A_ULP_ ## reg ## _ULIMIT) ulp_region(RX_ISCSI); ulp_region(RX_TDDP); ulp_region(TX_TPT); ulp_region(RX_STAG); ulp_region(RX_RQ); ulp_region(RX_RQUDP); ulp_region(RX_PBL); ulp_region(TX_PBL); #undef ulp_region md->base = 0; md->idx = nitems(region); if (!is_t4(sc) && t4_read_reg(sc, A_SGE_CONTROL2) & F_VFIFO_ENABLE) { md->base = G_BASEADDR(t4_read_reg(sc, A_SGE_DBVFIFO_BADDR)); md->limit = md->base + (G_DBVFIFO_SIZE((t4_read_reg(sc, A_SGE_DBVFIFO_SIZE))) << 2) - 1; } md++; md->base = t4_read_reg(sc, A_ULP_RX_CTX_BASE); md->limit = md->base + sc->tids.ntids - 1; md++; md->base = t4_read_reg(sc, A_ULP_TX_ERR_TABLE_BASE); md->limit = md->base + sc->tids.ntids - 1; md++; md->base = sc->vres.ocq.start; if (sc->vres.ocq.size) md->limit = md->base + sc->vres.ocq.size - 1; else md->idx = nitems(region); /* hide it */ md++; /* add any address-space holes, there can be up to 3 */ for (n = 0; n < i - 1; n++) if (avail[n].limit < avail[n + 1].base) (md++)->base = avail[n].limit; if (avail[n].limit) (md++)->base = avail[n].limit; n = md - mem; qsort(mem, n, sizeof(struct mem_desc), mem_desc_cmp); for (lo = 0; lo < i; lo++) mem_region_show(sb, memory[avail[lo].idx], avail[lo].base, avail[lo].limit - 1); sbuf_printf(sb, "\n"); for (i = 0; i < n; i++) { if (mem[i].idx >= nitems(region)) continue; /* skip holes */ if (!mem[i].limit) mem[i].limit = i < n - 1 ? mem[i + 1].base - 1 : ~0; mem_region_show(sb, region[mem[i].idx], mem[i].base, mem[i].limit); } sbuf_printf(sb, "\n"); lo = t4_read_reg(sc, A_CIM_SDRAM_BASE_ADDR); hi = t4_read_reg(sc, A_CIM_SDRAM_ADDR_SIZE) + lo - 1; mem_region_show(sb, "uP RAM:", lo, hi); lo = t4_read_reg(sc, A_CIM_EXTMEM2_BASE_ADDR); hi = t4_read_reg(sc, A_CIM_EXTMEM2_ADDR_SIZE) + lo - 1; mem_region_show(sb, "uP Extmem2:", lo, hi); lo = t4_read_reg(sc, A_TP_PMM_RX_MAX_PAGE); sbuf_printf(sb, "\n%u Rx pages of size %uKiB for %u channels\n", G_PMRXMAXPAGE(lo), t4_read_reg(sc, A_TP_PMM_RX_PAGE_SIZE) >> 10, (lo & F_PMRXNUMCHN) ? 2 : 1); lo = t4_read_reg(sc, A_TP_PMM_TX_MAX_PAGE); hi = t4_read_reg(sc, A_TP_PMM_TX_PAGE_SIZE); sbuf_printf(sb, "%u Tx pages of size %u%ciB for %u channels\n", G_PMTXMAXPAGE(lo), hi >= (1 << 20) ? (hi >> 20) : (hi >> 10), hi >= (1 << 20) ? 'M' : 'K', 1 << G_PMTXNUMCHN(lo)); sbuf_printf(sb, "%u p-structs\n", t4_read_reg(sc, A_TP_CMM_MM_MAX_PSTRUCT)); for (i = 0; i < 4; i++) { lo = t4_read_reg(sc, A_MPS_RX_PG_RSV0 + i * 4); if (is_t4(sc)) { used = G_USED(lo); alloc = G_ALLOC(lo); } else { used = G_T5_USED(lo); alloc = G_T5_ALLOC(lo); } sbuf_printf(sb, "\nPort %d using %u pages out of %u allocated", i, used, alloc); } for (i = 0; i < 4; i++) { lo = t4_read_reg(sc, A_MPS_RX_PG_RSV4 + i * 4); if (is_t4(sc)) { used = G_USED(lo); alloc = G_ALLOC(lo); } else { used = G_T5_USED(lo); alloc = G_T5_ALLOC(lo); } sbuf_printf(sb, "\nLoopback %d using %u pages out of %u allocated", i, used, alloc); } rc = sbuf_finish(sb); sbuf_delete(sb); return (rc); } static inline void tcamxy2valmask(uint64_t x, uint64_t y, uint8_t *addr, uint64_t *mask) { *mask = x | y; y = htobe64(y); memcpy(addr, (char *)&y + 2, ETHER_ADDR_LEN); } static int sysctl_mps_tcam(SYSCTL_HANDLER_ARGS) { struct adapter *sc = arg1; struct sbuf *sb; int rc, i, n; rc = sysctl_wire_old_buffer(req, 0); if (rc != 0) return (rc); sb = sbuf_new_for_sysctl(NULL, NULL, 4096, req); if (sb == NULL) return (ENOMEM); sbuf_printf(sb, "Idx Ethernet address Mask Vld Ports PF" " VF Replication P0 P1 P2 P3 ML"); n = is_t4(sc) ? NUM_MPS_CLS_SRAM_L_INSTANCES : NUM_MPS_T5_CLS_SRAM_L_INSTANCES; for (i = 0; i < n; i++) { uint64_t tcamx, tcamy, mask; uint32_t cls_lo, cls_hi; uint8_t addr[ETHER_ADDR_LEN]; tcamy = t4_read_reg64(sc, MPS_CLS_TCAM_Y_L(i)); tcamx = t4_read_reg64(sc, MPS_CLS_TCAM_X_L(i)); cls_lo = t4_read_reg(sc, MPS_CLS_SRAM_L(i)); cls_hi = t4_read_reg(sc, MPS_CLS_SRAM_H(i)); if (tcamx & tcamy) continue; tcamxy2valmask(tcamx, tcamy, addr, &mask); sbuf_printf(sb, "\n%3u %02x:%02x:%02x:%02x:%02x:%02x %012jx" " %c %#x%4u%4d", i, addr[0], addr[1], addr[2], addr[3], addr[4], addr[5], (uintmax_t)mask, (cls_lo & F_SRAM_VLD) ? 'Y' : 'N', G_PORTMAP(cls_hi), G_PF(cls_lo), (cls_lo & F_VF_VALID) ? G_VF(cls_lo) : -1); if (cls_lo & F_REPLICATE) { struct fw_ldst_cmd ldst_cmd; memset(&ldst_cmd, 0, sizeof(ldst_cmd)); ldst_cmd.op_to_addrspace = htobe32(V_FW_CMD_OP(FW_LDST_CMD) | F_FW_CMD_REQUEST | F_FW_CMD_READ | V_FW_LDST_CMD_ADDRSPACE(FW_LDST_ADDRSPC_MPS)); ldst_cmd.cycles_to_len16 = htobe32(FW_LEN16(ldst_cmd)); ldst_cmd.u.mps.fid_ctl = htobe16(V_FW_LDST_CMD_FID(FW_LDST_MPS_RPLC) | V_FW_LDST_CMD_CTL(i)); rc = begin_synchronized_op(sc, NULL, SLEEP_OK | INTR_OK, "t4mps"); if (rc) break; rc = -t4_wr_mbox(sc, sc->mbox, &ldst_cmd, sizeof(ldst_cmd), &ldst_cmd); end_synchronized_op(sc, 0); if (rc != 0) { sbuf_printf(sb, " ------------ error %3u ------------", rc); rc = 0; } else { sbuf_printf(sb, " %08x %08x %08x %08x", be32toh(ldst_cmd.u.mps.rplc127_96), be32toh(ldst_cmd.u.mps.rplc95_64), be32toh(ldst_cmd.u.mps.rplc63_32), be32toh(ldst_cmd.u.mps.rplc31_0)); } } else sbuf_printf(sb, "%36s", ""); sbuf_printf(sb, "%4u%3u%3u%3u %#3x", G_SRAM_PRIO0(cls_lo), G_SRAM_PRIO1(cls_lo), G_SRAM_PRIO2(cls_lo), G_SRAM_PRIO3(cls_lo), (cls_lo >> S_MULTILISTEN0) & 0xf); } if (rc) (void) sbuf_finish(sb); else rc = sbuf_finish(sb); sbuf_delete(sb); return (rc); } static int sysctl_path_mtus(SYSCTL_HANDLER_ARGS) { struct adapter *sc = arg1; struct sbuf *sb; int rc; uint16_t mtus[NMTUS]; rc = sysctl_wire_old_buffer(req, 0); if (rc != 0) return (rc); sb = sbuf_new_for_sysctl(NULL, NULL, 256, req); if (sb == NULL) return (ENOMEM); t4_read_mtu_tbl(sc, mtus, NULL); sbuf_printf(sb, "%u %u %u %u %u %u %u %u %u %u %u %u %u %u %u %u", mtus[0], mtus[1], mtus[2], mtus[3], mtus[4], mtus[5], mtus[6], mtus[7], mtus[8], mtus[9], mtus[10], mtus[11], mtus[12], mtus[13], mtus[14], mtus[15]); rc = sbuf_finish(sb); sbuf_delete(sb); return (rc); } static int sysctl_pm_stats(SYSCTL_HANDLER_ARGS) { struct adapter *sc = arg1; struct sbuf *sb; int rc, i; uint32_t cnt[PM_NSTATS]; uint64_t cyc[PM_NSTATS]; static const char *rx_stats[] = { "Read:", "Write bypass:", "Write mem:", "Flush:" }; static const char *tx_stats[] = { "Read:", "Write bypass:", "Write mem:", "Bypass + mem:" }; rc = sysctl_wire_old_buffer(req, 0); if (rc != 0) return (rc); sb = sbuf_new_for_sysctl(NULL, NULL, 256, req); if (sb == NULL) return (ENOMEM); t4_pmtx_get_stats(sc, cnt, cyc); sbuf_printf(sb, " Tx pcmds Tx bytes"); for (i = 0; i < ARRAY_SIZE(tx_stats); i++) sbuf_printf(sb, "\n%-13s %10u %20ju", tx_stats[i], cnt[i], cyc[i]); t4_pmrx_get_stats(sc, cnt, cyc); sbuf_printf(sb, "\n Rx pcmds Rx bytes"); for (i = 0; i < ARRAY_SIZE(rx_stats); i++) sbuf_printf(sb, "\n%-13s %10u %20ju", rx_stats[i], cnt[i], cyc[i]); rc = sbuf_finish(sb); sbuf_delete(sb); return (rc); } static int sysctl_rdma_stats(SYSCTL_HANDLER_ARGS) { struct adapter *sc = arg1; struct sbuf *sb; int rc; struct tp_rdma_stats stats; rc = sysctl_wire_old_buffer(req, 0); if (rc != 0) return (rc); sb = sbuf_new_for_sysctl(NULL, NULL, 256, req); if (sb == NULL) return (ENOMEM); t4_tp_get_rdma_stats(sc, &stats); sbuf_printf(sb, "NoRQEModDefferals: %u\n", stats.rqe_dfr_mod); sbuf_printf(sb, "NoRQEPktDefferals: %u", stats.rqe_dfr_pkt); rc = sbuf_finish(sb); sbuf_delete(sb); return (rc); } static int sysctl_tcp_stats(SYSCTL_HANDLER_ARGS) { struct adapter *sc = arg1; struct sbuf *sb; int rc; struct tp_tcp_stats v4, v6; rc = sysctl_wire_old_buffer(req, 0); if (rc != 0) return (rc); sb = sbuf_new_for_sysctl(NULL, NULL, 256, req); if (sb == NULL) return (ENOMEM); t4_tp_get_tcp_stats(sc, &v4, &v6); sbuf_printf(sb, " IP IPv6\n"); sbuf_printf(sb, "OutRsts: %20u %20u\n", v4.tcpOutRsts, v6.tcpOutRsts); sbuf_printf(sb, "InSegs: %20ju %20ju\n", v4.tcpInSegs, v6.tcpInSegs); sbuf_printf(sb, "OutSegs: %20ju %20ju\n", v4.tcpOutSegs, v6.tcpOutSegs); sbuf_printf(sb, "RetransSegs: %20ju %20ju", v4.tcpRetransSegs, v6.tcpRetransSegs); rc = sbuf_finish(sb); sbuf_delete(sb); return (rc); } static int sysctl_tids(SYSCTL_HANDLER_ARGS) { struct adapter *sc = arg1; struct sbuf *sb; int rc; struct tid_info *t = &sc->tids; rc = sysctl_wire_old_buffer(req, 0); if (rc != 0) return (rc); sb = sbuf_new_for_sysctl(NULL, NULL, 256, req); if (sb == NULL) return (ENOMEM); if (t->natids) { sbuf_printf(sb, "ATID range: 0-%u, in use: %u\n", t->natids - 1, t->atids_in_use); } if (t->ntids) { if (t4_read_reg(sc, A_LE_DB_CONFIG) & F_HASHEN) { uint32_t b = t4_read_reg(sc, A_LE_DB_SERVER_INDEX) / 4; if (b) { sbuf_printf(sb, "TID range: 0-%u, %u-%u", b - 1, t4_read_reg(sc, A_LE_DB_TID_HASHBASE) / 4, t->ntids - 1); } else { sbuf_printf(sb, "TID range: %u-%u", t4_read_reg(sc, A_LE_DB_TID_HASHBASE) / 4, t->ntids - 1); } } else sbuf_printf(sb, "TID range: 0-%u", t->ntids - 1); sbuf_printf(sb, ", in use: %u\n", atomic_load_acq_int(&t->tids_in_use)); } if (t->nstids) { sbuf_printf(sb, "STID range: %u-%u, in use: %u\n", t->stid_base, t->stid_base + t->nstids - 1, t->stids_in_use); } if (t->nftids) { sbuf_printf(sb, "FTID range: %u-%u\n", t->ftid_base, t->ftid_base + t->nftids - 1); } if (t->netids) { sbuf_printf(sb, "ETID range: %u-%u\n", t->etid_base, t->etid_base + t->netids - 1); } sbuf_printf(sb, "HW TID usage: %u IP users, %u IPv6 users", t4_read_reg(sc, A_LE_DB_ACT_CNT_IPV4), t4_read_reg(sc, A_LE_DB_ACT_CNT_IPV6)); rc = sbuf_finish(sb); sbuf_delete(sb); return (rc); } static int sysctl_tp_err_stats(SYSCTL_HANDLER_ARGS) { struct adapter *sc = arg1; struct sbuf *sb; int rc; struct tp_err_stats stats; rc = sysctl_wire_old_buffer(req, 0); if (rc != 0) return (rc); sb = sbuf_new_for_sysctl(NULL, NULL, 256, req); if (sb == NULL) return (ENOMEM); t4_tp_get_err_stats(sc, &stats); sbuf_printf(sb, " channel 0 channel 1 channel 2 " "channel 3\n"); sbuf_printf(sb, "macInErrs: %10u %10u %10u %10u\n", stats.macInErrs[0], stats.macInErrs[1], stats.macInErrs[2], stats.macInErrs[3]); sbuf_printf(sb, "hdrInErrs: %10u %10u %10u %10u\n", stats.hdrInErrs[0], stats.hdrInErrs[1], stats.hdrInErrs[2], stats.hdrInErrs[3]); sbuf_printf(sb, "tcpInErrs: %10u %10u %10u %10u\n", stats.tcpInErrs[0], stats.tcpInErrs[1], stats.tcpInErrs[2], stats.tcpInErrs[3]); sbuf_printf(sb, "tcp6InErrs: %10u %10u %10u %10u\n", stats.tcp6InErrs[0], stats.tcp6InErrs[1], stats.tcp6InErrs[2], stats.tcp6InErrs[3]); sbuf_printf(sb, "tnlCongDrops: %10u %10u %10u %10u\n", stats.tnlCongDrops[0], stats.tnlCongDrops[1], stats.tnlCongDrops[2], stats.tnlCongDrops[3]); sbuf_printf(sb, "tnlTxDrops: %10u %10u %10u %10u\n", stats.tnlTxDrops[0], stats.tnlTxDrops[1], stats.tnlTxDrops[2], stats.tnlTxDrops[3]); sbuf_printf(sb, "ofldVlanDrops: %10u %10u %10u %10u\n", stats.ofldVlanDrops[0], stats.ofldVlanDrops[1], stats.ofldVlanDrops[2], stats.ofldVlanDrops[3]); sbuf_printf(sb, "ofldChanDrops: %10u %10u %10u %10u\n\n", stats.ofldChanDrops[0], stats.ofldChanDrops[1], stats.ofldChanDrops[2], stats.ofldChanDrops[3]); sbuf_printf(sb, "ofldNoNeigh: %u\nofldCongDefer: %u", stats.ofldNoNeigh, stats.ofldCongDefer); rc = sbuf_finish(sb); sbuf_delete(sb); return (rc); } struct field_desc { const char *name; u_int start; u_int width; }; static void field_desc_show(struct sbuf *sb, uint64_t v, const struct field_desc *f) { char buf[32]; int line_size = 0; while (f->name) { uint64_t mask = (1ULL << f->width) - 1; int len = snprintf(buf, sizeof(buf), "%s: %ju", f->name, ((uintmax_t)v >> f->start) & mask); if (line_size + len >= 79) { line_size = 8; sbuf_printf(sb, "\n "); } sbuf_printf(sb, "%s ", buf); line_size += len + 1; f++; } sbuf_printf(sb, "\n"); } static struct field_desc tp_la0[] = { { "RcfOpCodeOut", 60, 4 }, { "State", 56, 4 }, { "WcfState", 52, 4 }, { "RcfOpcSrcOut", 50, 2 }, { "CRxError", 49, 1 }, { "ERxError", 48, 1 }, { "SanityFailed", 47, 1 }, { "SpuriousMsg", 46, 1 }, { "FlushInputMsg", 45, 1 }, { "FlushInputCpl", 44, 1 }, { "RssUpBit", 43, 1 }, { "RssFilterHit", 42, 1 }, { "Tid", 32, 10 }, { "InitTcb", 31, 1 }, { "LineNumber", 24, 7 }, { "Emsg", 23, 1 }, { "EdataOut", 22, 1 }, { "Cmsg", 21, 1 }, { "CdataOut", 20, 1 }, { "EreadPdu", 19, 1 }, { "CreadPdu", 18, 1 }, { "TunnelPkt", 17, 1 }, { "RcfPeerFin", 16, 1 }, { "RcfReasonOut", 12, 4 }, { "TxCchannel", 10, 2 }, { "RcfTxChannel", 8, 2 }, { "RxEchannel", 6, 2 }, { "RcfRxChannel", 5, 1 }, { "RcfDataOutSrdy", 4, 1 }, { "RxDvld", 3, 1 }, { "RxOoDvld", 2, 1 }, { "RxCongestion", 1, 1 }, { "TxCongestion", 0, 1 }, { NULL } }; static struct field_desc tp_la1[] = { { "CplCmdIn", 56, 8 }, { "CplCmdOut", 48, 8 }, { "ESynOut", 47, 1 }, { "EAckOut", 46, 1 }, { "EFinOut", 45, 1 }, { "ERstOut", 44, 1 }, { "SynIn", 43, 1 }, { "AckIn", 42, 1 }, { "FinIn", 41, 1 }, { "RstIn", 40, 1 }, { "DataIn", 39, 1 }, { "DataInVld", 38, 1 }, { "PadIn", 37, 1 }, { "RxBufEmpty", 36, 1 }, { "RxDdp", 35, 1 }, { "RxFbCongestion", 34, 1 }, { "TxFbCongestion", 33, 1 }, { "TxPktSumSrdy", 32, 1 }, { "RcfUlpType", 28, 4 }, { "Eread", 27, 1 }, { "Ebypass", 26, 1 }, { "Esave", 25, 1 }, { "Static0", 24, 1 }, { "Cread", 23, 1 }, { "Cbypass", 22, 1 }, { "Csave", 21, 1 }, { "CPktOut", 20, 1 }, { "RxPagePoolFull", 18, 2 }, { "RxLpbkPkt", 17, 1 }, { "TxLpbkPkt", 16, 1 }, { "RxVfValid", 15, 1 }, { "SynLearned", 14, 1 }, { "SetDelEntry", 13, 1 }, { "SetInvEntry", 12, 1 }, { "CpcmdDvld", 11, 1 }, { "CpcmdSave", 10, 1 }, { "RxPstructsFull", 8, 2 }, { "EpcmdDvld", 7, 1 }, { "EpcmdFlush", 6, 1 }, { "EpcmdTrimPrefix", 5, 1 }, { "EpcmdTrimPostfix", 4, 1 }, { "ERssIp4Pkt", 3, 1 }, { "ERssIp6Pkt", 2, 1 }, { "ERssTcpUdpPkt", 1, 1 }, { "ERssFceFipPkt", 0, 1 }, { NULL } }; static struct field_desc tp_la2[] = { { "CplCmdIn", 56, 8 }, { "MpsVfVld", 55, 1 }, { "MpsPf", 52, 3 }, { "MpsVf", 44, 8 }, { "SynIn", 43, 1 }, { "AckIn", 42, 1 }, { "FinIn", 41, 1 }, { "RstIn", 40, 1 }, { "DataIn", 39, 1 }, { "DataInVld", 38, 1 }, { "PadIn", 37, 1 }, { "RxBufEmpty", 36, 1 }, { "RxDdp", 35, 1 }, { "RxFbCongestion", 34, 1 }, { "TxFbCongestion", 33, 1 }, { "TxPktSumSrdy", 32, 1 }, { "RcfUlpType", 28, 4 }, { "Eread", 27, 1 }, { "Ebypass", 26, 1 }, { "Esave", 25, 1 }, { "Static0", 24, 1 }, { "Cread", 23, 1 }, { "Cbypass", 22, 1 }, { "Csave", 21, 1 }, { "CPktOut", 20, 1 }, { "RxPagePoolFull", 18, 2 }, { "RxLpbkPkt", 17, 1 }, { "TxLpbkPkt", 16, 1 }, { "RxVfValid", 15, 1 }, { "SynLearned", 14, 1 }, { "SetDelEntry", 13, 1 }, { "SetInvEntry", 12, 1 }, { "CpcmdDvld", 11, 1 }, { "CpcmdSave", 10, 1 }, { "RxPstructsFull", 8, 2 }, { "EpcmdDvld", 7, 1 }, { "EpcmdFlush", 6, 1 }, { "EpcmdTrimPrefix", 5, 1 }, { "EpcmdTrimPostfix", 4, 1 }, { "ERssIp4Pkt", 3, 1 }, { "ERssIp6Pkt", 2, 1 }, { "ERssTcpUdpPkt", 1, 1 }, { "ERssFceFipPkt", 0, 1 }, { NULL } }; static void tp_la_show(struct sbuf *sb, uint64_t *p, int idx) { field_desc_show(sb, *p, tp_la0); } static void tp_la_show2(struct sbuf *sb, uint64_t *p, int idx) { if (idx) sbuf_printf(sb, "\n"); field_desc_show(sb, p[0], tp_la0); if (idx < (TPLA_SIZE / 2 - 1) || p[1] != ~0ULL) field_desc_show(sb, p[1], tp_la0); } static void tp_la_show3(struct sbuf *sb, uint64_t *p, int idx) { if (idx) sbuf_printf(sb, "\n"); field_desc_show(sb, p[0], tp_la0); if (idx < (TPLA_SIZE / 2 - 1) || p[1] != ~0ULL) field_desc_show(sb, p[1], (p[0] & (1 << 17)) ? tp_la2 : tp_la1); } static int sysctl_tp_la(SYSCTL_HANDLER_ARGS) { struct adapter *sc = arg1; struct sbuf *sb; uint64_t *buf, *p; int rc; u_int i, inc; void (*show_func)(struct sbuf *, uint64_t *, int); rc = sysctl_wire_old_buffer(req, 0); if (rc != 0) return (rc); sb = sbuf_new_for_sysctl(NULL, NULL, 4096, req); if (sb == NULL) return (ENOMEM); buf = malloc(TPLA_SIZE * sizeof(uint64_t), M_CXGBE, M_ZERO | M_WAITOK); t4_tp_read_la(sc, buf, NULL); p = buf; switch (G_DBGLAMODE(t4_read_reg(sc, A_TP_DBG_LA_CONFIG))) { case 2: inc = 2; show_func = tp_la_show2; break; case 3: inc = 2; show_func = tp_la_show3; break; default: inc = 1; show_func = tp_la_show; } for (i = 0; i < TPLA_SIZE / inc; i++, p += inc) (*show_func)(sb, p, i); rc = sbuf_finish(sb); sbuf_delete(sb); free(buf, M_CXGBE); return (rc); } static int sysctl_tx_rate(SYSCTL_HANDLER_ARGS) { struct adapter *sc = arg1; struct sbuf *sb; int rc; u64 nrate[NCHAN], orate[NCHAN]; rc = sysctl_wire_old_buffer(req, 0); if (rc != 0) return (rc); sb = sbuf_new_for_sysctl(NULL, NULL, 256, req); if (sb == NULL) return (ENOMEM); t4_get_chan_txrate(sc, nrate, orate); sbuf_printf(sb, " channel 0 channel 1 channel 2 " "channel 3\n"); sbuf_printf(sb, "NIC B/s: %10ju %10ju %10ju %10ju\n", nrate[0], nrate[1], nrate[2], nrate[3]); sbuf_printf(sb, "Offload B/s: %10ju %10ju %10ju %10ju", orate[0], orate[1], orate[2], orate[3]); rc = sbuf_finish(sb); sbuf_delete(sb); return (rc); } static int sysctl_ulprx_la(SYSCTL_HANDLER_ARGS) { struct adapter *sc = arg1; struct sbuf *sb; uint32_t *buf, *p; int rc, i; rc = sysctl_wire_old_buffer(req, 0); if (rc != 0) return (rc); sb = sbuf_new_for_sysctl(NULL, NULL, 4096, req); if (sb == NULL) return (ENOMEM); buf = malloc(ULPRX_LA_SIZE * 8 * sizeof(uint32_t), M_CXGBE, M_ZERO | M_WAITOK); t4_ulprx_read_la(sc, buf); p = buf; sbuf_printf(sb, " Pcmd Type Message" " Data"); for (i = 0; i < ULPRX_LA_SIZE; i++, p += 8) { sbuf_printf(sb, "\n%08x%08x %4x %08x %08x%08x%08x%08x", p[1], p[0], p[2], p[3], p[7], p[6], p[5], p[4]); } rc = sbuf_finish(sb); sbuf_delete(sb); free(buf, M_CXGBE); return (rc); } static int sysctl_wcwr_stats(SYSCTL_HANDLER_ARGS) { struct adapter *sc = arg1; struct sbuf *sb; int rc, v; rc = sysctl_wire_old_buffer(req, 0); if (rc != 0) return (rc); sb = sbuf_new_for_sysctl(NULL, NULL, 4096, req); if (sb == NULL) return (ENOMEM); v = t4_read_reg(sc, A_SGE_STAT_CFG); if (G_STATSOURCE_T5(v) == 7) { if (G_STATMODE(v) == 0) { sbuf_printf(sb, "total %d, incomplete %d", t4_read_reg(sc, A_SGE_STAT_TOTAL), t4_read_reg(sc, A_SGE_STAT_MATCH)); } else if (G_STATMODE(v) == 1) { sbuf_printf(sb, "total %d, data overflow %d", t4_read_reg(sc, A_SGE_STAT_TOTAL), t4_read_reg(sc, A_SGE_STAT_MATCH)); } } rc = sbuf_finish(sb); sbuf_delete(sb); return (rc); } #endif static uint32_t fconf_to_mode(uint32_t fconf) { uint32_t mode; mode = T4_FILTER_IPv4 | T4_FILTER_IPv6 | T4_FILTER_IP_SADDR | T4_FILTER_IP_DADDR | T4_FILTER_IP_SPORT | T4_FILTER_IP_DPORT; if (fconf & F_FRAGMENTATION) mode |= T4_FILTER_IP_FRAGMENT; if (fconf & F_MPSHITTYPE) mode |= T4_FILTER_MPS_HIT_TYPE; if (fconf & F_MACMATCH) mode |= T4_FILTER_MAC_IDX; if (fconf & F_ETHERTYPE) mode |= T4_FILTER_ETH_TYPE; if (fconf & F_PROTOCOL) mode |= T4_FILTER_IP_PROTO; if (fconf & F_TOS) mode |= T4_FILTER_IP_TOS; if (fconf & F_VLAN) mode |= T4_FILTER_VLAN; if (fconf & F_VNIC_ID) mode |= T4_FILTER_VNIC; if (fconf & F_PORT) mode |= T4_FILTER_PORT; if (fconf & F_FCOE) mode |= T4_FILTER_FCoE; return (mode); } static uint32_t mode_to_fconf(uint32_t mode) { uint32_t fconf = 0; if (mode & T4_FILTER_IP_FRAGMENT) fconf |= F_FRAGMENTATION; if (mode & T4_FILTER_MPS_HIT_TYPE) fconf |= F_MPSHITTYPE; if (mode & T4_FILTER_MAC_IDX) fconf |= F_MACMATCH; if (mode & T4_FILTER_ETH_TYPE) fconf |= F_ETHERTYPE; if (mode & T4_FILTER_IP_PROTO) fconf |= F_PROTOCOL; if (mode & T4_FILTER_IP_TOS) fconf |= F_TOS; if (mode & T4_FILTER_VLAN) fconf |= F_VLAN; if (mode & T4_FILTER_VNIC) fconf |= F_VNIC_ID; if (mode & T4_FILTER_PORT) fconf |= F_PORT; if (mode & T4_FILTER_FCoE) fconf |= F_FCOE; return (fconf); } static uint32_t fspec_to_fconf(struct t4_filter_specification *fs) { uint32_t fconf = 0; if (fs->val.frag || fs->mask.frag) fconf |= F_FRAGMENTATION; if (fs->val.matchtype || fs->mask.matchtype) fconf |= F_MPSHITTYPE; if (fs->val.macidx || fs->mask.macidx) fconf |= F_MACMATCH; if (fs->val.ethtype || fs->mask.ethtype) fconf |= F_ETHERTYPE; if (fs->val.proto || fs->mask.proto) fconf |= F_PROTOCOL; if (fs->val.tos || fs->mask.tos) fconf |= F_TOS; if (fs->val.vlan_vld || fs->mask.vlan_vld) fconf |= F_VLAN; if (fs->val.vnic_vld || fs->mask.vnic_vld) fconf |= F_VNIC_ID; if (fs->val.iport || fs->mask.iport) fconf |= F_PORT; if (fs->val.fcoe || fs->mask.fcoe) fconf |= F_FCOE; return (fconf); } static int get_filter_mode(struct adapter *sc, uint32_t *mode) { int rc; uint32_t fconf; rc = begin_synchronized_op(sc, NULL, HOLD_LOCK | SLEEP_OK | INTR_OK, "t4getfm"); if (rc) return (rc); t4_read_indirect(sc, A_TP_PIO_ADDR, A_TP_PIO_DATA, &fconf, 1, A_TP_VLAN_PRI_MAP); if (sc->params.tp.vlan_pri_map != fconf) { log(LOG_WARNING, "%s: cached filter mode out of sync %x %x.\n", device_get_nameunit(sc->dev), sc->params.tp.vlan_pri_map, fconf); } *mode = fconf_to_mode(fconf); end_synchronized_op(sc, LOCK_HELD); return (0); } static int set_filter_mode(struct adapter *sc, uint32_t mode) { uint32_t fconf; int rc; fconf = mode_to_fconf(mode); rc = begin_synchronized_op(sc, NULL, HOLD_LOCK | SLEEP_OK | INTR_OK, "t4setfm"); if (rc) return (rc); if (sc->tids.ftids_in_use > 0) { rc = EBUSY; goto done; } #ifdef TCP_OFFLOAD if (uld_active(sc, ULD_TOM)) { rc = EBUSY; goto done; } #endif rc = -t4_set_filter_mode(sc, fconf); done: end_synchronized_op(sc, LOCK_HELD); return (rc); } static inline uint64_t get_filter_hits(struct adapter *sc, uint32_t fid) { uint32_t mw_base, off, tcb_base = t4_read_reg(sc, A_TP_CMM_TCB_BASE); uint64_t hits; memwin_info(sc, 0, &mw_base, NULL); off = position_memwin(sc, 0, tcb_base + (fid + sc->tids.ftid_base) * TCB_SIZE); if (is_t4(sc)) { hits = t4_read_reg64(sc, mw_base + off + 16); hits = be64toh(hits); } else { hits = t4_read_reg(sc, mw_base + off + 24); hits = be32toh(hits); } return (hits); } static int get_filter(struct adapter *sc, struct t4_filter *t) { int i, rc, nfilters = sc->tids.nftids; struct filter_entry *f; rc = begin_synchronized_op(sc, NULL, HOLD_LOCK | SLEEP_OK | INTR_OK, "t4getf"); if (rc) return (rc); if (sc->tids.ftids_in_use == 0 || sc->tids.ftid_tab == NULL || t->idx >= nfilters) { t->idx = 0xffffffff; goto done; } f = &sc->tids.ftid_tab[t->idx]; for (i = t->idx; i < nfilters; i++, f++) { if (f->valid) { t->idx = i; t->l2tidx = f->l2t ? f->l2t->idx : 0; t->smtidx = f->smtidx; if (f->fs.hitcnts) t->hits = get_filter_hits(sc, t->idx); else t->hits = UINT64_MAX; t->fs = f->fs; goto done; } } t->idx = 0xffffffff; done: end_synchronized_op(sc, LOCK_HELD); return (0); } static int set_filter(struct adapter *sc, struct t4_filter *t) { unsigned int nfilters, nports; struct filter_entry *f; int i, rc; rc = begin_synchronized_op(sc, NULL, SLEEP_OK | INTR_OK, "t4setf"); if (rc) return (rc); nfilters = sc->tids.nftids; nports = sc->params.nports; if (nfilters == 0) { rc = ENOTSUP; goto done; } if (!(sc->flags & FULL_INIT_DONE)) { rc = EAGAIN; goto done; } if (t->idx >= nfilters) { rc = EINVAL; goto done; } /* Validate against the global filter mode */ if ((sc->params.tp.vlan_pri_map | fspec_to_fconf(&t->fs)) != sc->params.tp.vlan_pri_map) { rc = E2BIG; goto done; } if (t->fs.action == FILTER_SWITCH && t->fs.eport >= nports) { rc = EINVAL; goto done; } if (t->fs.val.iport >= nports) { rc = EINVAL; goto done; } /* Can't specify an iq if not steering to it */ if (!t->fs.dirsteer && t->fs.iq) { rc = EINVAL; goto done; } /* IPv6 filter idx must be 4 aligned */ if (t->fs.type == 1 && ((t->idx & 0x3) || t->idx + 4 >= nfilters)) { rc = EINVAL; goto done; } if (sc->tids.ftid_tab == NULL) { KASSERT(sc->tids.ftids_in_use == 0, ("%s: no memory allocated but filters_in_use > 0", __func__)); sc->tids.ftid_tab = malloc(sizeof (struct filter_entry) * nfilters, M_CXGBE, M_NOWAIT | M_ZERO); if (sc->tids.ftid_tab == NULL) { rc = ENOMEM; goto done; } mtx_init(&sc->tids.ftid_lock, "T4 filters", 0, MTX_DEF); } for (i = 0; i < 4; i++) { f = &sc->tids.ftid_tab[t->idx + i]; if (f->pending || f->valid) { rc = EBUSY; goto done; } if (f->locked) { rc = EPERM; goto done; } if (t->fs.type == 0) break; } f = &sc->tids.ftid_tab[t->idx]; f->fs = t->fs; rc = set_filter_wr(sc, t->idx); done: end_synchronized_op(sc, 0); if (rc == 0) { mtx_lock(&sc->tids.ftid_lock); for (;;) { if (f->pending == 0) { rc = f->valid ? 0 : EIO; break; } if (mtx_sleep(&sc->tids.ftid_tab, &sc->tids.ftid_lock, PCATCH, "t4setfw", 0)) { rc = EINPROGRESS; break; } } mtx_unlock(&sc->tids.ftid_lock); } return (rc); } static int del_filter(struct adapter *sc, struct t4_filter *t) { unsigned int nfilters; struct filter_entry *f; int rc; rc = begin_synchronized_op(sc, NULL, SLEEP_OK | INTR_OK, "t4delf"); if (rc) return (rc); nfilters = sc->tids.nftids; if (nfilters == 0) { rc = ENOTSUP; goto done; } if (sc->tids.ftid_tab == NULL || sc->tids.ftids_in_use == 0 || t->idx >= nfilters) { rc = EINVAL; goto done; } if (!(sc->flags & FULL_INIT_DONE)) { rc = EAGAIN; goto done; } f = &sc->tids.ftid_tab[t->idx]; if (f->pending) { rc = EBUSY; goto done; } if (f->locked) { rc = EPERM; goto done; } if (f->valid) { t->fs = f->fs; /* extra info for the caller */ rc = del_filter_wr(sc, t->idx); } done: end_synchronized_op(sc, 0); if (rc == 0) { mtx_lock(&sc->tids.ftid_lock); for (;;) { if (f->pending == 0) { rc = f->valid ? EIO : 0; break; } if (mtx_sleep(&sc->tids.ftid_tab, &sc->tids.ftid_lock, PCATCH, "t4delfw", 0)) { rc = EINPROGRESS; break; } } mtx_unlock(&sc->tids.ftid_lock); } return (rc); } static void clear_filter(struct filter_entry *f) { if (f->l2t) t4_l2t_release(f->l2t); bzero(f, sizeof (*f)); } static int set_filter_wr(struct adapter *sc, int fidx) { struct filter_entry *f = &sc->tids.ftid_tab[fidx]; struct fw_filter_wr *fwr; unsigned int ftid; struct wrq_cookie cookie; ASSERT_SYNCHRONIZED_OP(sc); if (f->fs.newdmac || f->fs.newvlan) { /* This filter needs an L2T entry; allocate one. */ f->l2t = t4_l2t_alloc_switching(sc->l2t); if (f->l2t == NULL) return (EAGAIN); if (t4_l2t_set_switching(sc, f->l2t, f->fs.vlan, f->fs.eport, f->fs.dmac)) { t4_l2t_release(f->l2t); f->l2t = NULL; return (ENOMEM); } } ftid = sc->tids.ftid_base + fidx; fwr = start_wrq_wr(&sc->sge.mgmtq, howmany(sizeof(*fwr), 16), &cookie); if (fwr == NULL) return (ENOMEM); bzero(fwr, sizeof(*fwr)); fwr->op_pkd = htobe32(V_FW_WR_OP(FW_FILTER_WR)); fwr->len16_pkd = htobe32(FW_LEN16(*fwr)); fwr->tid_to_iq = htobe32(V_FW_FILTER_WR_TID(ftid) | V_FW_FILTER_WR_RQTYPE(f->fs.type) | V_FW_FILTER_WR_NOREPLY(0) | V_FW_FILTER_WR_IQ(f->fs.iq)); fwr->del_filter_to_l2tix = htobe32(V_FW_FILTER_WR_RPTTID(f->fs.rpttid) | V_FW_FILTER_WR_DROP(f->fs.action == FILTER_DROP) | V_FW_FILTER_WR_DIRSTEER(f->fs.dirsteer) | V_FW_FILTER_WR_MASKHASH(f->fs.maskhash) | V_FW_FILTER_WR_DIRSTEERHASH(f->fs.dirsteerhash) | V_FW_FILTER_WR_LPBK(f->fs.action == FILTER_SWITCH) | V_FW_FILTER_WR_DMAC(f->fs.newdmac) | V_FW_FILTER_WR_SMAC(f->fs.newsmac) | V_FW_FILTER_WR_INSVLAN(f->fs.newvlan == VLAN_INSERT || f->fs.newvlan == VLAN_REWRITE) | V_FW_FILTER_WR_RMVLAN(f->fs.newvlan == VLAN_REMOVE || f->fs.newvlan == VLAN_REWRITE) | V_FW_FILTER_WR_HITCNTS(f->fs.hitcnts) | V_FW_FILTER_WR_TXCHAN(f->fs.eport) | V_FW_FILTER_WR_PRIO(f->fs.prio) | V_FW_FILTER_WR_L2TIX(f->l2t ? f->l2t->idx : 0)); fwr->ethtype = htobe16(f->fs.val.ethtype); fwr->ethtypem = htobe16(f->fs.mask.ethtype); fwr->frag_to_ovlan_vldm = (V_FW_FILTER_WR_FRAG(f->fs.val.frag) | V_FW_FILTER_WR_FRAGM(f->fs.mask.frag) | V_FW_FILTER_WR_IVLAN_VLD(f->fs.val.vlan_vld) | V_FW_FILTER_WR_OVLAN_VLD(f->fs.val.vnic_vld) | V_FW_FILTER_WR_IVLAN_VLDM(f->fs.mask.vlan_vld) | V_FW_FILTER_WR_OVLAN_VLDM(f->fs.mask.vnic_vld)); fwr->smac_sel = 0; fwr->rx_chan_rx_rpl_iq = htobe16(V_FW_FILTER_WR_RX_CHAN(0) | V_FW_FILTER_WR_RX_RPL_IQ(sc->sge.fwq.abs_id)); fwr->maci_to_matchtypem = htobe32(V_FW_FILTER_WR_MACI(f->fs.val.macidx) | V_FW_FILTER_WR_MACIM(f->fs.mask.macidx) | V_FW_FILTER_WR_FCOE(f->fs.val.fcoe) | V_FW_FILTER_WR_FCOEM(f->fs.mask.fcoe) | V_FW_FILTER_WR_PORT(f->fs.val.iport) | V_FW_FILTER_WR_PORTM(f->fs.mask.iport) | V_FW_FILTER_WR_MATCHTYPE(f->fs.val.matchtype) | V_FW_FILTER_WR_MATCHTYPEM(f->fs.mask.matchtype)); fwr->ptcl = f->fs.val.proto; fwr->ptclm = f->fs.mask.proto; fwr->ttyp = f->fs.val.tos; fwr->ttypm = f->fs.mask.tos; fwr->ivlan = htobe16(f->fs.val.vlan); fwr->ivlanm = htobe16(f->fs.mask.vlan); fwr->ovlan = htobe16(f->fs.val.vnic); fwr->ovlanm = htobe16(f->fs.mask.vnic); bcopy(f->fs.val.dip, fwr->lip, sizeof (fwr->lip)); bcopy(f->fs.mask.dip, fwr->lipm, sizeof (fwr->lipm)); bcopy(f->fs.val.sip, fwr->fip, sizeof (fwr->fip)); bcopy(f->fs.mask.sip, fwr->fipm, sizeof (fwr->fipm)); fwr->lp = htobe16(f->fs.val.dport); fwr->lpm = htobe16(f->fs.mask.dport); fwr->fp = htobe16(f->fs.val.sport); fwr->fpm = htobe16(f->fs.mask.sport); if (f->fs.newsmac) bcopy(f->fs.smac, fwr->sma, sizeof (fwr->sma)); f->pending = 1; sc->tids.ftids_in_use++; commit_wrq_wr(&sc->sge.mgmtq, fwr, &cookie); return (0); } static int del_filter_wr(struct adapter *sc, int fidx) { struct filter_entry *f = &sc->tids.ftid_tab[fidx]; struct fw_filter_wr *fwr; unsigned int ftid; struct wrq_cookie cookie; ftid = sc->tids.ftid_base + fidx; fwr = start_wrq_wr(&sc->sge.mgmtq, howmany(sizeof(*fwr), 16), &cookie); if (fwr == NULL) return (ENOMEM); bzero(fwr, sizeof (*fwr)); t4_mk_filtdelwr(ftid, fwr, sc->sge.fwq.abs_id); f->pending = 1; commit_wrq_wr(&sc->sge.mgmtq, fwr, &cookie); return (0); } int t4_filter_rpl(struct sge_iq *iq, const struct rss_header *rss, struct mbuf *m) { struct adapter *sc = iq->adapter; const struct cpl_set_tcb_rpl *rpl = (const void *)(rss + 1); unsigned int idx = GET_TID(rpl); unsigned int rc; struct filter_entry *f; KASSERT(m == NULL, ("%s: payload with opcode %02x", __func__, rss->opcode)); if (is_ftid(sc, idx)) { idx -= sc->tids.ftid_base; f = &sc->tids.ftid_tab[idx]; rc = G_COOKIE(rpl->cookie); mtx_lock(&sc->tids.ftid_lock); if (rc == FW_FILTER_WR_FLT_ADDED) { KASSERT(f->pending, ("%s: filter[%u] isn't pending.", __func__, idx)); f->smtidx = (be64toh(rpl->oldval) >> 24) & 0xff; f->pending = 0; /* asynchronous setup completed */ f->valid = 1; } else { if (rc != FW_FILTER_WR_FLT_DELETED) { /* Add or delete failed, display an error */ log(LOG_ERR, "filter %u setup failed with error %u\n", idx, rc); } clear_filter(f); sc->tids.ftids_in_use--; } wakeup(&sc->tids.ftid_tab); mtx_unlock(&sc->tids.ftid_lock); } return (0); } static int get_sge_context(struct adapter *sc, struct t4_sge_context *cntxt) { int rc; if (cntxt->cid > M_CTXTQID) return (EINVAL); if (cntxt->mem_id != CTXT_EGRESS && cntxt->mem_id != CTXT_INGRESS && cntxt->mem_id != CTXT_FLM && cntxt->mem_id != CTXT_CNM) return (EINVAL); rc = begin_synchronized_op(sc, NULL, SLEEP_OK | INTR_OK, "t4ctxt"); if (rc) return (rc); if (sc->flags & FW_OK) { rc = -t4_sge_ctxt_rd(sc, sc->mbox, cntxt->cid, cntxt->mem_id, &cntxt->data[0]); if (rc == 0) goto done; } /* * Read via firmware failed or wasn't even attempted. Read directly via * the backdoor. */ rc = -t4_sge_ctxt_rd_bd(sc, cntxt->cid, cntxt->mem_id, &cntxt->data[0]); done: end_synchronized_op(sc, 0); return (rc); } static int load_fw(struct adapter *sc, struct t4_data *fw) { int rc; uint8_t *fw_data; rc = begin_synchronized_op(sc, NULL, SLEEP_OK | INTR_OK, "t4ldfw"); if (rc) return (rc); if (sc->flags & FULL_INIT_DONE) { rc = EBUSY; goto done; } fw_data = malloc(fw->len, M_CXGBE, M_WAITOK); if (fw_data == NULL) { rc = ENOMEM; goto done; } rc = copyin(fw->data, fw_data, fw->len); if (rc == 0) rc = -t4_load_fw(sc, fw_data, fw->len); free(fw_data, M_CXGBE); done: end_synchronized_op(sc, 0); return (rc); } static int read_card_mem(struct adapter *sc, int win, struct t4_mem_range *mr) { uint32_t addr, off, remaining, i, n; uint32_t *buf, *b; uint32_t mw_base, mw_aperture; int rc; uint8_t *dst; rc = validate_mem_range(sc, mr->addr, mr->len); if (rc != 0) return (rc); memwin_info(sc, win, &mw_base, &mw_aperture); buf = b = malloc(min(mr->len, mw_aperture), M_CXGBE, M_WAITOK); addr = mr->addr; remaining = mr->len; dst = (void *)mr->data; while (remaining) { off = position_memwin(sc, win, addr); /* number of bytes that we'll copy in the inner loop */ n = min(remaining, mw_aperture - off); for (i = 0; i < n; i += 4) *b++ = t4_read_reg(sc, mw_base + off + i); rc = copyout(buf, dst, n); if (rc != 0) break; b = buf; dst += n; remaining -= n; addr += n; } free(buf, M_CXGBE); return (rc); } static int read_i2c(struct adapter *sc, struct t4_i2c_data *i2cd) { int rc; if (i2cd->len == 0 || i2cd->port_id >= sc->params.nports) return (EINVAL); if (i2cd->len > sizeof(i2cd->data)) return (EFBIG); rc = begin_synchronized_op(sc, NULL, SLEEP_OK | INTR_OK, "t4i2crd"); if (rc) return (rc); rc = -t4_i2c_rd(sc, sc->mbox, i2cd->port_id, i2cd->dev_addr, i2cd->offset, i2cd->len, &i2cd->data[0]); end_synchronized_op(sc, 0); return (rc); } static int in_range(int val, int lo, int hi) { return (val < 0 || (val <= hi && val >= lo)); } static int set_sched_class(struct adapter *sc, struct t4_sched_params *p) { int fw_subcmd, fw_type, rc; rc = begin_synchronized_op(sc, NULL, SLEEP_OK | INTR_OK, "t4setsc"); if (rc) return (rc); if (!(sc->flags & FULL_INIT_DONE)) { rc = EAGAIN; goto done; } /* * Translate the cxgbetool parameters into T4 firmware parameters. (The * sub-command and type are in common locations.) */ if (p->subcmd == SCHED_CLASS_SUBCMD_CONFIG) fw_subcmd = FW_SCHED_SC_CONFIG; else if (p->subcmd == SCHED_CLASS_SUBCMD_PARAMS) fw_subcmd = FW_SCHED_SC_PARAMS; else { rc = EINVAL; goto done; } if (p->type == SCHED_CLASS_TYPE_PACKET) fw_type = FW_SCHED_TYPE_PKTSCHED; else { rc = EINVAL; goto done; } if (fw_subcmd == FW_SCHED_SC_CONFIG) { /* Vet our parameters ..*/ if (p->u.config.minmax < 0) { rc = EINVAL; goto done; } /* And pass the request to the firmware ...*/ rc = -t4_sched_config(sc, fw_type, p->u.config.minmax, 1); goto done; } if (fw_subcmd == FW_SCHED_SC_PARAMS) { int fw_level; int fw_mode; int fw_rateunit; int fw_ratemode; if (p->u.params.level == SCHED_CLASS_LEVEL_CL_RL) fw_level = FW_SCHED_PARAMS_LEVEL_CL_RL; else if (p->u.params.level == SCHED_CLASS_LEVEL_CL_WRR) fw_level = FW_SCHED_PARAMS_LEVEL_CL_WRR; else if (p->u.params.level == SCHED_CLASS_LEVEL_CH_RL) fw_level = FW_SCHED_PARAMS_LEVEL_CH_RL; else { rc = EINVAL; goto done; } if (p->u.params.mode == SCHED_CLASS_MODE_CLASS) fw_mode = FW_SCHED_PARAMS_MODE_CLASS; else if (p->u.params.mode == SCHED_CLASS_MODE_FLOW) fw_mode = FW_SCHED_PARAMS_MODE_FLOW; else { rc = EINVAL; goto done; } if (p->u.params.rateunit == SCHED_CLASS_RATEUNIT_BITS) fw_rateunit = FW_SCHED_PARAMS_UNIT_BITRATE; else if (p->u.params.rateunit == SCHED_CLASS_RATEUNIT_PKTS) fw_rateunit = FW_SCHED_PARAMS_UNIT_PKTRATE; else { rc = EINVAL; goto done; } if (p->u.params.ratemode == SCHED_CLASS_RATEMODE_REL) fw_ratemode = FW_SCHED_PARAMS_RATE_REL; else if (p->u.params.ratemode == SCHED_CLASS_RATEMODE_ABS) fw_ratemode = FW_SCHED_PARAMS_RATE_ABS; else { rc = EINVAL; goto done; } /* Vet our parameters ... */ if (!in_range(p->u.params.channel, 0, 3) || !in_range(p->u.params.cl, 0, is_t4(sc) ? 15 : 16) || !in_range(p->u.params.minrate, 0, 10000000) || !in_range(p->u.params.maxrate, 0, 10000000) || !in_range(p->u.params.weight, 0, 100)) { rc = ERANGE; goto done; } /* * Translate any unset parameters into the firmware's * nomenclature and/or fail the call if the parameters * are required ... */ if (p->u.params.rateunit < 0 || p->u.params.ratemode < 0 || p->u.params.channel < 0 || p->u.params.cl < 0) { rc = EINVAL; goto done; } if (p->u.params.minrate < 0) p->u.params.minrate = 0; if (p->u.params.maxrate < 0) { if (p->u.params.level == SCHED_CLASS_LEVEL_CL_RL || p->u.params.level == SCHED_CLASS_LEVEL_CH_RL) { rc = EINVAL; goto done; } else p->u.params.maxrate = 0; } if (p->u.params.weight < 0) { if (p->u.params.level == SCHED_CLASS_LEVEL_CL_WRR) { rc = EINVAL; goto done; } else p->u.params.weight = 0; } if (p->u.params.pktsize < 0) { if (p->u.params.level == SCHED_CLASS_LEVEL_CL_RL || p->u.params.level == SCHED_CLASS_LEVEL_CH_RL) { rc = EINVAL; goto done; } else p->u.params.pktsize = 0; } /* See what the firmware thinks of the request ... */ rc = -t4_sched_params(sc, fw_type, fw_level, fw_mode, fw_rateunit, fw_ratemode, p->u.params.channel, p->u.params.cl, p->u.params.minrate, p->u.params.maxrate, p->u.params.weight, p->u.params.pktsize, 1); goto done; } rc = EINVAL; done: end_synchronized_op(sc, 0); return (rc); } static int set_sched_queue(struct adapter *sc, struct t4_sched_queue *p) { struct port_info *pi = NULL; struct sge_txq *txq; uint32_t fw_mnem, fw_queue, fw_class; int i, rc; rc = begin_synchronized_op(sc, NULL, SLEEP_OK | INTR_OK, "t4setsq"); if (rc) return (rc); if (!(sc->flags & FULL_INIT_DONE)) { rc = EAGAIN; goto done; } if (p->port >= sc->params.nports) { rc = EINVAL; goto done; } pi = sc->port[p->port]; if (!in_range(p->queue, 0, pi->ntxq - 1) || !in_range(p->cl, 0, 7)) { rc = EINVAL; goto done; } /* * Create a template for the FW_PARAMS_CMD mnemonic and value (TX * Scheduling Class in this case). */ fw_mnem = (V_FW_PARAMS_MNEM(FW_PARAMS_MNEM_DMAQ) | V_FW_PARAMS_PARAM_X(FW_PARAMS_PARAM_DMAQ_EQ_SCHEDCLASS_ETH)); fw_class = p->cl < 0 ? 0xffffffff : p->cl; /* * If op.queue is non-negative, then we're only changing the scheduling * on a single specified TX queue. */ if (p->queue >= 0) { txq = &sc->sge.txq[pi->first_txq + p->queue]; fw_queue = (fw_mnem | V_FW_PARAMS_PARAM_YZ(txq->eq.cntxt_id)); rc = -t4_set_params(sc, sc->mbox, sc->pf, 0, 1, &fw_queue, &fw_class); goto done; } /* * Change the scheduling on all the TX queues for the * interface. */ for_each_txq(pi, i, txq) { fw_queue = (fw_mnem | V_FW_PARAMS_PARAM_YZ(txq->eq.cntxt_id)); rc = -t4_set_params(sc, sc->mbox, sc->pf, 0, 1, &fw_queue, &fw_class); if (rc) goto done; } rc = 0; done: end_synchronized_op(sc, 0); return (rc); } int t4_os_find_pci_capability(struct adapter *sc, int cap) { int i; return (pci_find_cap(sc->dev, cap, &i) == 0 ? i : 0); } int t4_os_pci_save_state(struct adapter *sc) { device_t dev; struct pci_devinfo *dinfo; dev = sc->dev; dinfo = device_get_ivars(dev); pci_cfg_save(dev, dinfo, 0); return (0); } int t4_os_pci_restore_state(struct adapter *sc) { device_t dev; struct pci_devinfo *dinfo; dev = sc->dev; dinfo = device_get_ivars(dev); pci_cfg_restore(dev, dinfo); return (0); } void t4_os_portmod_changed(const struct adapter *sc, int idx) { struct port_info *pi = sc->port[idx]; static const char *mod_str[] = { NULL, "LR", "SR", "ER", "TWINAX", "active TWINAX", "LRM" }; build_medialist(pi, &pi->media); #ifdef DEV_NETMAP build_medialist(pi, &pi->nm_media); #endif if (pi->mod_type == FW_PORT_MOD_TYPE_NONE) if_printf(pi->ifp, "transceiver unplugged.\n"); else if (pi->mod_type == FW_PORT_MOD_TYPE_UNKNOWN) if_printf(pi->ifp, "unknown transceiver inserted.\n"); else if (pi->mod_type == FW_PORT_MOD_TYPE_NOTSUPPORTED) if_printf(pi->ifp, "unsupported transceiver inserted.\n"); else if (pi->mod_type > 0 && pi->mod_type < nitems(mod_str)) { if_printf(pi->ifp, "%s transceiver inserted.\n", mod_str[pi->mod_type]); } else { if_printf(pi->ifp, "transceiver (type %d) inserted.\n", pi->mod_type); } } void t4_os_link_changed(struct adapter *sc, int idx, int link_stat, int reason) { struct port_info *pi = sc->port[idx]; struct ifnet *ifp = pi->ifp; if (link_stat) { pi->linkdnrc = -1; ifp->if_baudrate = IF_Mbps(pi->link_cfg.speed); if_link_state_change(ifp, LINK_STATE_UP); } else { if (reason >= 0) pi->linkdnrc = reason; if_link_state_change(ifp, LINK_STATE_DOWN); } } void t4_iterate(void (*func)(struct adapter *, void *), void *arg) { struct adapter *sc; sx_slock(&t4_list_lock); SLIST_FOREACH(sc, &t4_list, link) { /* * func should not make any assumptions about what state sc is * in - the only guarantee is that sc->sc_lock is a valid lock. */ func(sc, arg); } sx_sunlock(&t4_list_lock); } static int t4_open(struct cdev *dev, int flags, int type, struct thread *td) { return (0); } static int t4_close(struct cdev *dev, int flags, int type, struct thread *td) { return (0); } static int t4_ioctl(struct cdev *dev, unsigned long cmd, caddr_t data, int fflag, struct thread *td) { int rc; struct adapter *sc = dev->si_drv1; rc = priv_check(td, PRIV_DRIVER); if (rc != 0) return (rc); switch (cmd) { case CHELSIO_T4_GETREG: { struct t4_reg *edata = (struct t4_reg *)data; if ((edata->addr & 0x3) != 0 || edata->addr >= sc->mmio_len) return (EFAULT); if (edata->size == 4) edata->val = t4_read_reg(sc, edata->addr); else if (edata->size == 8) edata->val = t4_read_reg64(sc, edata->addr); else return (EINVAL); break; } case CHELSIO_T4_SETREG: { struct t4_reg *edata = (struct t4_reg *)data; if ((edata->addr & 0x3) != 0 || edata->addr >= sc->mmio_len) return (EFAULT); if (edata->size == 4) { if (edata->val & 0xffffffff00000000) return (EINVAL); t4_write_reg(sc, edata->addr, (uint32_t) edata->val); } else if (edata->size == 8) t4_write_reg64(sc, edata->addr, edata->val); else return (EINVAL); break; } case CHELSIO_T4_REGDUMP: { struct t4_regdump *regs = (struct t4_regdump *)data; int reglen = is_t4(sc) ? T4_REGDUMP_SIZE : T5_REGDUMP_SIZE; uint8_t *buf; if (regs->len < reglen) { regs->len = reglen; /* hint to the caller */ return (ENOBUFS); } regs->len = reglen; buf = malloc(reglen, M_CXGBE, M_WAITOK | M_ZERO); t4_get_regs(sc, regs, buf); rc = copyout(buf, regs->data, reglen); free(buf, M_CXGBE); break; } case CHELSIO_T4_GET_FILTER_MODE: rc = get_filter_mode(sc, (uint32_t *)data); break; case CHELSIO_T4_SET_FILTER_MODE: rc = set_filter_mode(sc, *(uint32_t *)data); break; case CHELSIO_T4_GET_FILTER: rc = get_filter(sc, (struct t4_filter *)data); break; case CHELSIO_T4_SET_FILTER: rc = set_filter(sc, (struct t4_filter *)data); break; case CHELSIO_T4_DEL_FILTER: rc = del_filter(sc, (struct t4_filter *)data); break; case CHELSIO_T4_GET_SGE_CONTEXT: rc = get_sge_context(sc, (struct t4_sge_context *)data); break; case CHELSIO_T4_LOAD_FW: rc = load_fw(sc, (struct t4_data *)data); break; case CHELSIO_T4_GET_MEM: rc = read_card_mem(sc, 2, (struct t4_mem_range *)data); break; case CHELSIO_T4_GET_I2C: rc = read_i2c(sc, (struct t4_i2c_data *)data); break; case CHELSIO_T4_CLEAR_STATS: { int i; u_int port_id = *(uint32_t *)data; struct port_info *pi; if (port_id >= sc->params.nports) return (EINVAL); pi = sc->port[port_id]; /* MAC stats */ t4_clr_port_stats(sc, pi->tx_chan); pi->tx_parse_error = 0; if (pi->flags & PORT_INIT_DONE) { struct sge_rxq *rxq; struct sge_txq *txq; struct sge_wrq *wrq; for_each_rxq(pi, i, rxq) { #if defined(INET) || defined(INET6) rxq->lro.lro_queued = 0; rxq->lro.lro_flushed = 0; #endif rxq->rxcsum = 0; rxq->vlan_extraction = 0; } for_each_txq(pi, i, txq) { txq->txcsum = 0; txq->tso_wrs = 0; txq->vlan_insertion = 0; txq->imm_wrs = 0; txq->sgl_wrs = 0; txq->txpkt_wrs = 0; txq->txpkts0_wrs = 0; txq->txpkts1_wrs = 0; txq->txpkts0_pkts = 0; txq->txpkts1_pkts = 0; mp_ring_reset_stats(txq->r); } #ifdef TCP_OFFLOAD /* nothing to clear for each ofld_rxq */ for_each_ofld_txq(pi, i, wrq) { wrq->tx_wrs_direct = 0; wrq->tx_wrs_copied = 0; } #endif wrq = &sc->sge.ctrlq[pi->port_id]; wrq->tx_wrs_direct = 0; wrq->tx_wrs_copied = 0; } break; } case CHELSIO_T4_SCHED_CLASS: rc = set_sched_class(sc, (struct t4_sched_params *)data); break; case CHELSIO_T4_SCHED_QUEUE: rc = set_sched_queue(sc, (struct t4_sched_queue *)data); break; case CHELSIO_T4_GET_TRACER: rc = t4_get_tracer(sc, (struct t4_tracer *)data); break; case CHELSIO_T4_SET_TRACER: rc = t4_set_tracer(sc, (struct t4_tracer *)data); break; default: rc = EINVAL; } return (rc); } #ifdef TCP_OFFLOAD void t4_iscsi_init(struct ifnet *ifp, unsigned int tag_mask, const unsigned int *pgsz_order) { struct port_info *pi = ifp->if_softc; struct adapter *sc = pi->adapter; t4_write_reg(sc, A_ULP_RX_ISCSI_TAGMASK, tag_mask); t4_write_reg(sc, A_ULP_RX_ISCSI_PSZ, V_HPZ0(pgsz_order[0]) | V_HPZ1(pgsz_order[1]) | V_HPZ2(pgsz_order[2]) | V_HPZ3(pgsz_order[3])); } static int toe_capability(struct port_info *pi, int enable) { int rc; struct adapter *sc = pi->adapter; ASSERT_SYNCHRONIZED_OP(sc); if (!is_offload(sc)) return (ENODEV); if (enable) { /* * We need the port's queues around so that we're able to send * and receive CPLs to/from the TOE even if the ifnet for this * port has never been UP'd administratively. */ if (!(pi->flags & PORT_INIT_DONE)) { rc = cxgbe_init_synchronized(pi); if (rc) return (rc); } if (isset(&sc->offload_map, pi->port_id)) return (0); if (!uld_active(sc, ULD_TOM)) { rc = t4_activate_uld(sc, ULD_TOM); if (rc == EAGAIN) { log(LOG_WARNING, "You must kldload t4_tom.ko before trying " "to enable TOE on a cxgbe interface.\n"); } if (rc != 0) return (rc); KASSERT(sc->tom_softc != NULL, ("%s: TOM activated but softc NULL", __func__)); KASSERT(uld_active(sc, ULD_TOM), ("%s: TOM activated but flag not set", __func__)); } /* Activate iWARP and iSCSI too, if the modules are loaded. */ if (!uld_active(sc, ULD_IWARP)) (void) t4_activate_uld(sc, ULD_IWARP); if (!uld_active(sc, ULD_ISCSI)) (void) t4_activate_uld(sc, ULD_ISCSI); setbit(&sc->offload_map, pi->port_id); } else { if (!isset(&sc->offload_map, pi->port_id)) return (0); KASSERT(uld_active(sc, ULD_TOM), ("%s: TOM never initialized?", __func__)); clrbit(&sc->offload_map, pi->port_id); } return (0); } /* * Add an upper layer driver to the global list. */ int t4_register_uld(struct uld_info *ui) { int rc = 0; struct uld_info *u; sx_xlock(&t4_uld_list_lock); SLIST_FOREACH(u, &t4_uld_list, link) { if (u->uld_id == ui->uld_id) { rc = EEXIST; goto done; } } SLIST_INSERT_HEAD(&t4_uld_list, ui, link); ui->refcount = 0; done: sx_xunlock(&t4_uld_list_lock); return (rc); } int t4_unregister_uld(struct uld_info *ui) { int rc = EINVAL; struct uld_info *u; sx_xlock(&t4_uld_list_lock); SLIST_FOREACH(u, &t4_uld_list, link) { if (u == ui) { if (ui->refcount > 0) { rc = EBUSY; goto done; } SLIST_REMOVE(&t4_uld_list, ui, uld_info, link); rc = 0; goto done; } } done: sx_xunlock(&t4_uld_list_lock); return (rc); } int t4_activate_uld(struct adapter *sc, int id) { int rc; struct uld_info *ui; ASSERT_SYNCHRONIZED_OP(sc); if (id < 0 || id > ULD_MAX) return (EINVAL); rc = EAGAIN; /* kldoad the module with this ULD and try again. */ sx_slock(&t4_uld_list_lock); SLIST_FOREACH(ui, &t4_uld_list, link) { if (ui->uld_id == id) { if (!(sc->flags & FULL_INIT_DONE)) { rc = adapter_full_init(sc); if (rc != 0) break; } rc = ui->activate(sc); if (rc == 0) { setbit(&sc->active_ulds, id); ui->refcount++; } break; } } sx_sunlock(&t4_uld_list_lock); return (rc); } int t4_deactivate_uld(struct adapter *sc, int id) { int rc; struct uld_info *ui; ASSERT_SYNCHRONIZED_OP(sc); if (id < 0 || id > ULD_MAX) return (EINVAL); rc = ENXIO; sx_slock(&t4_uld_list_lock); SLIST_FOREACH(ui, &t4_uld_list, link) { if (ui->uld_id == id) { rc = ui->deactivate(sc); if (rc == 0) { clrbit(&sc->active_ulds, id); ui->refcount--; } break; } } sx_sunlock(&t4_uld_list_lock); return (rc); } int uld_active(struct adapter *sc, int uld_id) { MPASS(uld_id >= 0 && uld_id <= ULD_MAX); return (isset(&sc->active_ulds, uld_id)); } #endif /* * Come up with reasonable defaults for some of the tunables, provided they're * not set by the user (in which case we'll use the values as is). */ static void tweak_tunables(void) { int nc = mp_ncpus; /* our snapshot of the number of CPUs */ if (t4_ntxq10g < 1) t4_ntxq10g = min(nc, NTXQ_10G); if (t4_ntxq1g < 1) t4_ntxq1g = min(nc, NTXQ_1G); if (t4_nrxq10g < 1) t4_nrxq10g = min(nc, NRXQ_10G); if (t4_nrxq1g < 1) t4_nrxq1g = min(nc, NRXQ_1G); #ifdef TCP_OFFLOAD if (t4_nofldtxq10g < 1) t4_nofldtxq10g = min(nc, NOFLDTXQ_10G); if (t4_nofldtxq1g < 1) t4_nofldtxq1g = min(nc, NOFLDTXQ_1G); if (t4_nofldrxq10g < 1) t4_nofldrxq10g = min(nc, NOFLDRXQ_10G); if (t4_nofldrxq1g < 1) t4_nofldrxq1g = min(nc, NOFLDRXQ_1G); if (t4_toecaps_allowed == -1) t4_toecaps_allowed = FW_CAPS_CONFIG_TOE; #else if (t4_toecaps_allowed == -1) t4_toecaps_allowed = 0; #endif #ifdef DEV_NETMAP if (t4_nnmtxq10g < 1) t4_nnmtxq10g = min(nc, NNMTXQ_10G); if (t4_nnmtxq1g < 1) t4_nnmtxq1g = min(nc, NNMTXQ_1G); if (t4_nnmrxq10g < 1) t4_nnmrxq10g = min(nc, NNMRXQ_10G); if (t4_nnmrxq1g < 1) t4_nnmrxq1g = min(nc, NNMRXQ_1G); #endif if (t4_tmr_idx_10g < 0 || t4_tmr_idx_10g >= SGE_NTIMERS) t4_tmr_idx_10g = TMR_IDX_10G; if (t4_pktc_idx_10g < -1 || t4_pktc_idx_10g >= SGE_NCOUNTERS) t4_pktc_idx_10g = PKTC_IDX_10G; if (t4_tmr_idx_1g < 0 || t4_tmr_idx_1g >= SGE_NTIMERS) t4_tmr_idx_1g = TMR_IDX_1G; if (t4_pktc_idx_1g < -1 || t4_pktc_idx_1g >= SGE_NCOUNTERS) t4_pktc_idx_1g = PKTC_IDX_1G; if (t4_qsize_txq < 128) t4_qsize_txq = 128; if (t4_qsize_rxq < 128) t4_qsize_rxq = 128; while (t4_qsize_rxq & 7) t4_qsize_rxq++; t4_intr_types &= INTR_MSIX | INTR_MSI | INTR_INTX; } static struct sx mlu; /* mod load unload */ SX_SYSINIT(cxgbe_mlu, &mlu, "cxgbe mod load/unload"); static int mod_event(module_t mod, int cmd, void *arg) { int rc = 0; static int loaded = 0; switch (cmd) { case MOD_LOAD: sx_xlock(&mlu); if (loaded++ == 0) { t4_sge_modload(); sx_init(&t4_list_lock, "T4/T5 adapters"); SLIST_INIT(&t4_list); #ifdef TCP_OFFLOAD sx_init(&t4_uld_list_lock, "T4/T5 ULDs"); SLIST_INIT(&t4_uld_list); #endif t4_tracer_modload(); tweak_tunables(); } sx_xunlock(&mlu); break; case MOD_UNLOAD: sx_xlock(&mlu); if (--loaded == 0) { int tries; sx_slock(&t4_list_lock); if (!SLIST_EMPTY(&t4_list)) { rc = EBUSY; sx_sunlock(&t4_list_lock); goto done_unload; } #ifdef TCP_OFFLOAD sx_slock(&t4_uld_list_lock); if (!SLIST_EMPTY(&t4_uld_list)) { rc = EBUSY; sx_sunlock(&t4_uld_list_lock); sx_sunlock(&t4_list_lock); goto done_unload; } #endif tries = 0; while (tries++ < 5 && t4_sge_extfree_refs() != 0) { uprintf("%ju clusters with custom free routine " "still is use.\n", t4_sge_extfree_refs()); pause("t4unload", 2 * hz); } #ifdef TCP_OFFLOAD sx_sunlock(&t4_uld_list_lock); #endif sx_sunlock(&t4_list_lock); if (t4_sge_extfree_refs() == 0) { t4_tracer_modunload(); #ifdef TCP_OFFLOAD sx_destroy(&t4_uld_list_lock); #endif sx_destroy(&t4_list_lock); t4_sge_modunload(); loaded = 0; } else { rc = EBUSY; loaded++; /* undo earlier decrement */ } } done_unload: sx_xunlock(&mlu); break; } return (rc); } static devclass_t t4_devclass, t5_devclass; static devclass_t cxgbe_devclass, cxl_devclass; DRIVER_MODULE(t4nex, pci, t4_driver, t4_devclass, mod_event, 0); MODULE_VERSION(t4nex, 1); MODULE_DEPEND(t4nex, firmware, 1, 1, 1); DRIVER_MODULE(t5nex, pci, t5_driver, t5_devclass, mod_event, 0); MODULE_VERSION(t5nex, 1); MODULE_DEPEND(t5nex, firmware, 1, 1, 1); DRIVER_MODULE(cxgbe, t4nex, cxgbe_driver, cxgbe_devclass, 0, 0); MODULE_VERSION(cxgbe, 1); DRIVER_MODULE(cxl, t5nex, cxl_driver, cxl_devclass, 0, 0); MODULE_VERSION(cxl, 1); Index: projects/clang-trunk/sys/dev/cxgbe/t4_sge.c =================================================================== --- projects/clang-trunk/sys/dev/cxgbe/t4_sge.c (revision 283870) +++ projects/clang-trunk/sys/dev/cxgbe/t4_sge.c (revision 283871) @@ -1,4774 +1,4774 @@ /*- * Copyright (c) 2011 Chelsio Communications, Inc. * All rights reserved. * Written by: Navdeep Parhar * * Redistribution and use in source and binary forms, with or without * modification, are permitted provided that the following conditions * are met: * 1. Redistributions of source code must retain the above copyright * notice, this list of conditions and the following disclaimer. * 2. Redistributions in binary form must reproduce the above copyright * notice, this list of conditions and the following disclaimer in the * documentation and/or other materials provided with the distribution. * * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE * ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF * SUCH DAMAGE. */ #include __FBSDID("$FreeBSD$"); #include "opt_inet.h" #include "opt_inet6.h" #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #ifdef DEV_NETMAP #include #include #include #include #include #endif #include "common/common.h" #include "common/t4_regs.h" #include "common/t4_regs_values.h" #include "common/t4_msg.h" #include "t4_mp_ring.h" #ifdef T4_PKT_TIMESTAMP #define RX_COPY_THRESHOLD (MINCLSIZE - 8) #else #define RX_COPY_THRESHOLD MINCLSIZE #endif /* * Ethernet frames are DMA'd at this byte offset into the freelist buffer. * 0-7 are valid values. */ int fl_pktshift = 2; TUNABLE_INT("hw.cxgbe.fl_pktshift", &fl_pktshift); /* * Pad ethernet payload up to this boundary. * -1: driver should figure out a good value. * 0: disable padding. * Any power of 2 from 32 to 4096 (both inclusive) is also a valid value. */ int fl_pad = -1; TUNABLE_INT("hw.cxgbe.fl_pad", &fl_pad); /* * Status page length. * -1: driver should figure out a good value. * 64 or 128 are the only other valid values. */ int spg_len = -1; TUNABLE_INT("hw.cxgbe.spg_len", &spg_len); /* * Congestion drops. * -1: no congestion feedback (not recommended). * 0: backpressure the channel instead of dropping packets right away. * 1: no backpressure, drop packets for the congested queue immediately. */ static int cong_drop = 0; TUNABLE_INT("hw.cxgbe.cong_drop", &cong_drop); /* * Deliver multiple frames in the same free list buffer if they fit. * -1: let the driver decide whether to enable buffer packing or not. * 0: disable buffer packing. * 1: enable buffer packing. */ static int buffer_packing = -1; TUNABLE_INT("hw.cxgbe.buffer_packing", &buffer_packing); /* * Start next frame in a packed buffer at this boundary. * -1: driver should figure out a good value. * T4: driver will ignore this and use the same value as fl_pad above. * T5: 16, or a power of 2 from 64 to 4096 (both inclusive) is a valid value. */ static int fl_pack = -1; TUNABLE_INT("hw.cxgbe.fl_pack", &fl_pack); /* * Allow the driver to create mbuf(s) in a cluster allocated for rx. * 0: never; always allocate mbufs from the zone_mbuf UMA zone. * 1: ok to create mbuf(s) within a cluster if there is room. */ static int allow_mbufs_in_cluster = 1; TUNABLE_INT("hw.cxgbe.allow_mbufs_in_cluster", &allow_mbufs_in_cluster); /* * Largest rx cluster size that the driver is allowed to allocate. */ static int largest_rx_cluster = MJUM16BYTES; TUNABLE_INT("hw.cxgbe.largest_rx_cluster", &largest_rx_cluster); /* * Size of cluster allocation that's most likely to succeed. The driver will * fall back to this size if it fails to allocate clusters larger than this. */ static int safest_rx_cluster = PAGE_SIZE; TUNABLE_INT("hw.cxgbe.safest_rx_cluster", &safest_rx_cluster); struct txpkts { u_int wr_type; /* type 0 or type 1 */ u_int npkt; /* # of packets in this work request */ u_int plen; /* total payload (sum of all packets) */ u_int len16; /* # of 16B pieces used by this work request */ }; /* A packet's SGL. This + m_pkthdr has all info needed for tx */ struct sgl { struct sglist sg; struct sglist_seg seg[TX_SGL_SEGS]; }; static int service_iq(struct sge_iq *, int); static struct mbuf *get_fl_payload(struct adapter *, struct sge_fl *, uint32_t); static int t4_eth_rx(struct sge_iq *, const struct rss_header *, struct mbuf *); static inline void init_iq(struct sge_iq *, struct adapter *, int, int, int); static inline void init_fl(struct adapter *, struct sge_fl *, int, int, char *); static inline void init_eq(struct sge_eq *, int, int, uint8_t, uint16_t, char *); static int alloc_ring(struct adapter *, size_t, bus_dma_tag_t *, bus_dmamap_t *, bus_addr_t *, void **); static int free_ring(struct adapter *, bus_dma_tag_t, bus_dmamap_t, bus_addr_t, void *); static int alloc_iq_fl(struct port_info *, struct sge_iq *, struct sge_fl *, int, int); static int free_iq_fl(struct port_info *, struct sge_iq *, struct sge_fl *); static void add_fl_sysctls(struct sysctl_ctx_list *, struct sysctl_oid *, struct sge_fl *); static int alloc_fwq(struct adapter *); static int free_fwq(struct adapter *); static int alloc_mgmtq(struct adapter *); static int free_mgmtq(struct adapter *); static int alloc_rxq(struct port_info *, struct sge_rxq *, int, int, struct sysctl_oid *); static int free_rxq(struct port_info *, struct sge_rxq *); #ifdef TCP_OFFLOAD static int alloc_ofld_rxq(struct port_info *, struct sge_ofld_rxq *, int, int, struct sysctl_oid *); static int free_ofld_rxq(struct port_info *, struct sge_ofld_rxq *); #endif #ifdef DEV_NETMAP static int alloc_nm_rxq(struct port_info *, struct sge_nm_rxq *, int, int, struct sysctl_oid *); static int free_nm_rxq(struct port_info *, struct sge_nm_rxq *); static int alloc_nm_txq(struct port_info *, struct sge_nm_txq *, int, int, struct sysctl_oid *); static int free_nm_txq(struct port_info *, struct sge_nm_txq *); #endif static int ctrl_eq_alloc(struct adapter *, struct sge_eq *); static int eth_eq_alloc(struct adapter *, struct port_info *, struct sge_eq *); #ifdef TCP_OFFLOAD static int ofld_eq_alloc(struct adapter *, struct port_info *, struct sge_eq *); #endif static int alloc_eq(struct adapter *, struct port_info *, struct sge_eq *); static int free_eq(struct adapter *, struct sge_eq *); static int alloc_wrq(struct adapter *, struct port_info *, struct sge_wrq *, struct sysctl_oid *); static int free_wrq(struct adapter *, struct sge_wrq *); static int alloc_txq(struct port_info *, struct sge_txq *, int, struct sysctl_oid *); static int free_txq(struct port_info *, struct sge_txq *); static void oneseg_dma_callback(void *, bus_dma_segment_t *, int, int); static inline void ring_fl_db(struct adapter *, struct sge_fl *); static int refill_fl(struct adapter *, struct sge_fl *, int); static void refill_sfl(void *); static int alloc_fl_sdesc(struct sge_fl *); static void free_fl_sdesc(struct adapter *, struct sge_fl *); static void find_best_refill_source(struct adapter *, struct sge_fl *, int); static void find_safe_refill_source(struct adapter *, struct sge_fl *); static void add_fl_to_sfl(struct adapter *, struct sge_fl *); static inline void get_pkt_gl(struct mbuf *, struct sglist *); static inline u_int txpkt_len16(u_int, u_int); static inline u_int txpkts0_len16(u_int); static inline u_int txpkts1_len16(void); static u_int write_txpkt_wr(struct sge_txq *, struct fw_eth_tx_pkt_wr *, struct mbuf *, u_int); static int try_txpkts(struct mbuf *, struct mbuf *, struct txpkts *, u_int); static int add_to_txpkts(struct mbuf *, struct txpkts *, u_int); static u_int write_txpkts_wr(struct sge_txq *, struct fw_eth_tx_pkts_wr *, struct mbuf *, const struct txpkts *, u_int); static void write_gl_to_txd(struct sge_txq *, struct mbuf *, caddr_t *, int); static inline void copy_to_txd(struct sge_eq *, caddr_t, caddr_t *, int); static inline void ring_eq_db(struct adapter *, struct sge_eq *, u_int); static inline uint16_t read_hw_cidx(struct sge_eq *); static inline u_int reclaimable_tx_desc(struct sge_eq *); static inline u_int total_available_tx_desc(struct sge_eq *); static u_int reclaim_tx_descs(struct sge_txq *, u_int); static void tx_reclaim(void *, int); static __be64 get_flit(struct sglist_seg *, int, int); static int handle_sge_egr_update(struct sge_iq *, const struct rss_header *, struct mbuf *); static int handle_fw_msg(struct sge_iq *, const struct rss_header *, struct mbuf *); static void wrq_tx_drain(void *, int); static void drain_wrq_wr_list(struct adapter *, struct sge_wrq *); static int sysctl_uint16(SYSCTL_HANDLER_ARGS); static int sysctl_bufsizes(SYSCTL_HANDLER_ARGS); static counter_u64_t extfree_refs; static counter_u64_t extfree_rels; /* * Called on MOD_LOAD. Validates and calculates the SGE tunables. */ void t4_sge_modload(void) { if (fl_pktshift < 0 || fl_pktshift > 7) { printf("Invalid hw.cxgbe.fl_pktshift value (%d)," " using 2 instead.\n", fl_pktshift); fl_pktshift = 2; } if (spg_len != 64 && spg_len != 128) { int len; #if defined(__i386__) || defined(__amd64__) len = cpu_clflush_line_size > 64 ? 128 : 64; #else len = 64; #endif if (spg_len != -1) { printf("Invalid hw.cxgbe.spg_len value (%d)," " using %d instead.\n", spg_len, len); } spg_len = len; } if (cong_drop < -1 || cong_drop > 1) { printf("Invalid hw.cxgbe.cong_drop value (%d)," " using 0 instead.\n", cong_drop); cong_drop = 0; } extfree_refs = counter_u64_alloc(M_WAITOK); extfree_rels = counter_u64_alloc(M_WAITOK); counter_u64_zero(extfree_refs); counter_u64_zero(extfree_rels); } void t4_sge_modunload(void) { counter_u64_free(extfree_refs); counter_u64_free(extfree_rels); } uint64_t t4_sge_extfree_refs(void) { uint64_t refs, rels; rels = counter_u64_fetch(extfree_rels); refs = counter_u64_fetch(extfree_refs); return (refs - rels); } void t4_init_sge_cpl_handlers(struct adapter *sc) { t4_register_cpl_handler(sc, CPL_FW4_MSG, handle_fw_msg); t4_register_cpl_handler(sc, CPL_FW6_MSG, handle_fw_msg); t4_register_cpl_handler(sc, CPL_SGE_EGR_UPDATE, handle_sge_egr_update); t4_register_cpl_handler(sc, CPL_RX_PKT, t4_eth_rx); t4_register_fw_msg_handler(sc, FW6_TYPE_CMD_RPL, t4_handle_fw_rpl); } static inline void setup_pad_and_pack_boundaries(struct adapter *sc) { uint32_t v, m; int pad, pack; pad = fl_pad; if (fl_pad < 32 || fl_pad > 4096 || !powerof2(fl_pad)) { /* * If there is any chance that we might use buffer packing and * the chip is a T4, then pick 64 as the pad/pack boundary. Set * it to 32 in all other cases. */ pad = is_t4(sc) && buffer_packing ? 64 : 32; /* * For fl_pad = 0 we'll still write a reasonable value to the * register but all the freelists will opt out of padding. * We'll complain here only if the user tried to set it to a * value greater than 0 that was invalid. */ if (fl_pad > 0) { device_printf(sc->dev, "Invalid hw.cxgbe.fl_pad value" " (%d), using %d instead.\n", fl_pad, pad); } } m = V_INGPADBOUNDARY(M_INGPADBOUNDARY); v = V_INGPADBOUNDARY(ilog2(pad) - 5); t4_set_reg_field(sc, A_SGE_CONTROL, m, v); if (is_t4(sc)) { if (fl_pack != -1 && fl_pack != pad) { /* Complain but carry on. */ device_printf(sc->dev, "hw.cxgbe.fl_pack (%d) ignored," " using %d instead.\n", fl_pack, pad); } return; } pack = fl_pack; if (fl_pack < 16 || fl_pack == 32 || fl_pack > 4096 || !powerof2(fl_pack)) { pack = max(sc->params.pci.mps, CACHE_LINE_SIZE); MPASS(powerof2(pack)); if (pack < 16) pack = 16; if (pack == 32) pack = 64; if (pack > 4096) pack = 4096; if (fl_pack != -1) { device_printf(sc->dev, "Invalid hw.cxgbe.fl_pack value" " (%d), using %d instead.\n", fl_pack, pack); } } m = V_INGPACKBOUNDARY(M_INGPACKBOUNDARY); if (pack == 16) v = V_INGPACKBOUNDARY(0); else v = V_INGPACKBOUNDARY(ilog2(pack) - 5); MPASS(!is_t4(sc)); /* T4 doesn't have SGE_CONTROL2 */ t4_set_reg_field(sc, A_SGE_CONTROL2, m, v); } /* * adap->params.vpd.cclk must be set up before this is called. */ void t4_tweak_chip_settings(struct adapter *sc) { int i; uint32_t v, m; int intr_timer[SGE_NTIMERS] = {1, 5, 10, 50, 100, 200}; int timer_max = M_TIMERVALUE0 * 1000 / sc->params.vpd.cclk; int intr_pktcount[SGE_NCOUNTERS] = {1, 8, 16, 32}; /* 63 max */ uint16_t indsz = min(RX_COPY_THRESHOLD - 1, M_INDICATESIZE); static int sge_flbuf_sizes[] = { MCLBYTES, #if MJUMPAGESIZE != MCLBYTES MJUMPAGESIZE, MJUMPAGESIZE - CL_METADATA_SIZE, MJUMPAGESIZE - 2 * MSIZE - CL_METADATA_SIZE, #endif MJUM9BYTES, MJUM16BYTES, MCLBYTES - MSIZE - CL_METADATA_SIZE, MJUM9BYTES - CL_METADATA_SIZE, MJUM16BYTES - CL_METADATA_SIZE, }; KASSERT(sc->flags & MASTER_PF, ("%s: trying to change chip settings when not master.", __func__)); m = V_PKTSHIFT(M_PKTSHIFT) | F_RXPKTCPLMODE | F_EGRSTATUSPAGESIZE; v = V_PKTSHIFT(fl_pktshift) | F_RXPKTCPLMODE | V_EGRSTATUSPAGESIZE(spg_len == 128); t4_set_reg_field(sc, A_SGE_CONTROL, m, v); setup_pad_and_pack_boundaries(sc); v = V_HOSTPAGESIZEPF0(PAGE_SHIFT - 10) | V_HOSTPAGESIZEPF1(PAGE_SHIFT - 10) | V_HOSTPAGESIZEPF2(PAGE_SHIFT - 10) | V_HOSTPAGESIZEPF3(PAGE_SHIFT - 10) | V_HOSTPAGESIZEPF4(PAGE_SHIFT - 10) | V_HOSTPAGESIZEPF5(PAGE_SHIFT - 10) | V_HOSTPAGESIZEPF6(PAGE_SHIFT - 10) | V_HOSTPAGESIZEPF7(PAGE_SHIFT - 10); t4_write_reg(sc, A_SGE_HOST_PAGE_SIZE, v); KASSERT(nitems(sge_flbuf_sizes) <= SGE_FLBUF_SIZES, ("%s: hw buffer size table too big", __func__)); for (i = 0; i < min(nitems(sge_flbuf_sizes), SGE_FLBUF_SIZES); i++) { t4_write_reg(sc, A_SGE_FL_BUFFER_SIZE0 + (4 * i), sge_flbuf_sizes[i]); } v = V_THRESHOLD_0(intr_pktcount[0]) | V_THRESHOLD_1(intr_pktcount[1]) | V_THRESHOLD_2(intr_pktcount[2]) | V_THRESHOLD_3(intr_pktcount[3]); t4_write_reg(sc, A_SGE_INGRESS_RX_THRESHOLD, v); KASSERT(intr_timer[0] <= timer_max, ("%s: not a single usable timer (%d, %d)", __func__, intr_timer[0], timer_max)); for (i = 1; i < nitems(intr_timer); i++) { KASSERT(intr_timer[i] >= intr_timer[i - 1], ("%s: timers not listed in increasing order (%d)", __func__, i)); while (intr_timer[i] > timer_max) { if (i == nitems(intr_timer) - 1) { intr_timer[i] = timer_max; break; } intr_timer[i] += intr_timer[i - 1]; intr_timer[i] /= 2; } } v = V_TIMERVALUE0(us_to_core_ticks(sc, intr_timer[0])) | V_TIMERVALUE1(us_to_core_ticks(sc, intr_timer[1])); t4_write_reg(sc, A_SGE_TIMER_VALUE_0_AND_1, v); v = V_TIMERVALUE2(us_to_core_ticks(sc, intr_timer[2])) | V_TIMERVALUE3(us_to_core_ticks(sc, intr_timer[3])); t4_write_reg(sc, A_SGE_TIMER_VALUE_2_AND_3, v); v = V_TIMERVALUE4(us_to_core_ticks(sc, intr_timer[4])) | V_TIMERVALUE5(us_to_core_ticks(sc, intr_timer[5])); t4_write_reg(sc, A_SGE_TIMER_VALUE_4_AND_5, v); if (cong_drop == 0) { m = F_TUNNELCNGDROP0 | F_TUNNELCNGDROP1 | F_TUNNELCNGDROP2 | F_TUNNELCNGDROP3; t4_set_reg_field(sc, A_TP_PARA_REG3, m, 0); } /* 4K, 16K, 64K, 256K DDP "page sizes" */ v = V_HPZ0(0) | V_HPZ1(2) | V_HPZ2(4) | V_HPZ3(6); t4_write_reg(sc, A_ULP_RX_TDDP_PSZ, v); m = v = F_TDDPTAGTCB; t4_set_reg_field(sc, A_ULP_RX_CTL, m, v); m = V_INDICATESIZE(M_INDICATESIZE) | F_REARMDDPOFFSET | F_RESETDDPOFFSET; v = V_INDICATESIZE(indsz) | F_REARMDDPOFFSET | F_RESETDDPOFFSET; t4_set_reg_field(sc, A_TP_PARA_REG5, m, v); } /* * SGE wants the buffer to be at least 64B and then a multiple of 16. If * padding is is use the buffer's start and end need to be aligned to the pad * boundary as well. We'll just make sure that the size is a multiple of the * boundary here, it is up to the buffer allocation code to make sure the start * of the buffer is aligned as well. */ static inline int hwsz_ok(struct adapter *sc, int hwsz) { int mask = fl_pad ? sc->sge.pad_boundary - 1 : 16 - 1; return (hwsz >= 64 && (hwsz & mask) == 0); } /* * XXX: driver really should be able to deal with unexpected settings. */ int t4_read_chip_settings(struct adapter *sc) { struct sge *s = &sc->sge; int i, j, n, rc = 0; uint32_t m, v, r; uint16_t indsz = min(RX_COPY_THRESHOLD - 1, M_INDICATESIZE); static int sw_buf_sizes[] = { /* Sorted by size */ MCLBYTES, #if MJUMPAGESIZE != MCLBYTES MJUMPAGESIZE, #endif MJUM9BYTES, MJUM16BYTES }; struct sw_zone_info *swz, *safe_swz; struct hw_buf_info *hwb; m = V_PKTSHIFT(M_PKTSHIFT) | F_RXPKTCPLMODE | F_EGRSTATUSPAGESIZE; v = V_PKTSHIFT(fl_pktshift) | F_RXPKTCPLMODE | V_EGRSTATUSPAGESIZE(spg_len == 128); r = t4_read_reg(sc, A_SGE_CONTROL); if ((r & m) != v) { device_printf(sc->dev, "invalid SGE_CONTROL(0x%x)\n", r); rc = EINVAL; } s->pad_boundary = 1 << (G_INGPADBOUNDARY(r) + 5); if (is_t4(sc)) s->pack_boundary = s->pad_boundary; else { r = t4_read_reg(sc, A_SGE_CONTROL2); if (G_INGPACKBOUNDARY(r) == 0) s->pack_boundary = 16; else s->pack_boundary = 1 << (G_INGPACKBOUNDARY(r) + 5); } v = V_HOSTPAGESIZEPF0(PAGE_SHIFT - 10) | V_HOSTPAGESIZEPF1(PAGE_SHIFT - 10) | V_HOSTPAGESIZEPF2(PAGE_SHIFT - 10) | V_HOSTPAGESIZEPF3(PAGE_SHIFT - 10) | V_HOSTPAGESIZEPF4(PAGE_SHIFT - 10) | V_HOSTPAGESIZEPF5(PAGE_SHIFT - 10) | V_HOSTPAGESIZEPF6(PAGE_SHIFT - 10) | V_HOSTPAGESIZEPF7(PAGE_SHIFT - 10); r = t4_read_reg(sc, A_SGE_HOST_PAGE_SIZE); if (r != v) { device_printf(sc->dev, "invalid SGE_HOST_PAGE_SIZE(0x%x)\n", r); rc = EINVAL; } /* Filter out unusable hw buffer sizes entirely (mark with -2). */ hwb = &s->hw_buf_info[0]; for (i = 0; i < nitems(s->hw_buf_info); i++, hwb++) { r = t4_read_reg(sc, A_SGE_FL_BUFFER_SIZE0 + (4 * i)); hwb->size = r; hwb->zidx = hwsz_ok(sc, r) ? -1 : -2; hwb->next = -1; } /* * Create a sorted list in decreasing order of hw buffer sizes (and so * increasing order of spare area) for each software zone. * * If padding is enabled then the start and end of the buffer must align * to the pad boundary; if packing is enabled then they must align with * the pack boundary as well. Allocations from the cluster zones are * aligned to min(size, 4K), so the buffer starts at that alignment and * ends at hwb->size alignment. If mbuf inlining is allowed the * starting alignment will be reduced to MSIZE and the driver will * exercise appropriate caution when deciding on the best buffer layout * to use. */ n = 0; /* no usable buffer size to begin with */ swz = &s->sw_zone_info[0]; safe_swz = NULL; for (i = 0; i < SW_ZONE_SIZES; i++, swz++) { int8_t head = -1, tail = -1; swz->size = sw_buf_sizes[i]; swz->zone = m_getzone(swz->size); swz->type = m_gettype(swz->size); if (swz->size < PAGE_SIZE) { MPASS(powerof2(swz->size)); if (fl_pad && (swz->size % sc->sge.pad_boundary != 0)) continue; } if (swz->size == safest_rx_cluster) safe_swz = swz; hwb = &s->hw_buf_info[0]; for (j = 0; j < SGE_FLBUF_SIZES; j++, hwb++) { if (hwb->zidx != -1 || hwb->size > swz->size) continue; #ifdef INVARIANTS if (fl_pad) MPASS(hwb->size % sc->sge.pad_boundary == 0); #endif hwb->zidx = i; if (head == -1) head = tail = j; else if (hwb->size < s->hw_buf_info[tail].size) { s->hw_buf_info[tail].next = j; tail = j; } else { int8_t *cur; struct hw_buf_info *t; for (cur = &head; *cur != -1; cur = &t->next) { t = &s->hw_buf_info[*cur]; if (hwb->size == t->size) { hwb->zidx = -2; break; } if (hwb->size > t->size) { hwb->next = *cur; *cur = j; break; } } } } swz->head_hwidx = head; swz->tail_hwidx = tail; if (tail != -1) { n++; if (swz->size - s->hw_buf_info[tail].size >= CL_METADATA_SIZE) sc->flags |= BUF_PACKING_OK; } } if (n == 0) { device_printf(sc->dev, "no usable SGE FL buffer size.\n"); rc = EINVAL; } s->safe_hwidx1 = -1; s->safe_hwidx2 = -1; if (safe_swz != NULL) { s->safe_hwidx1 = safe_swz->head_hwidx; for (i = safe_swz->head_hwidx; i != -1; i = hwb->next) { int spare; hwb = &s->hw_buf_info[i]; #ifdef INVARIANTS if (fl_pad) MPASS(hwb->size % sc->sge.pad_boundary == 0); #endif spare = safe_swz->size - hwb->size; if (spare >= CL_METADATA_SIZE) { s->safe_hwidx2 = i; break; } } } r = t4_read_reg(sc, A_SGE_INGRESS_RX_THRESHOLD); s->counter_val[0] = G_THRESHOLD_0(r); s->counter_val[1] = G_THRESHOLD_1(r); s->counter_val[2] = G_THRESHOLD_2(r); s->counter_val[3] = G_THRESHOLD_3(r); r = t4_read_reg(sc, A_SGE_TIMER_VALUE_0_AND_1); s->timer_val[0] = G_TIMERVALUE0(r) / core_ticks_per_usec(sc); s->timer_val[1] = G_TIMERVALUE1(r) / core_ticks_per_usec(sc); r = t4_read_reg(sc, A_SGE_TIMER_VALUE_2_AND_3); s->timer_val[2] = G_TIMERVALUE2(r) / core_ticks_per_usec(sc); s->timer_val[3] = G_TIMERVALUE3(r) / core_ticks_per_usec(sc); r = t4_read_reg(sc, A_SGE_TIMER_VALUE_4_AND_5); s->timer_val[4] = G_TIMERVALUE4(r) / core_ticks_per_usec(sc); s->timer_val[5] = G_TIMERVALUE5(r) / core_ticks_per_usec(sc); if (cong_drop == 0) { m = F_TUNNELCNGDROP0 | F_TUNNELCNGDROP1 | F_TUNNELCNGDROP2 | F_TUNNELCNGDROP3; r = t4_read_reg(sc, A_TP_PARA_REG3); if (r & m) { device_printf(sc->dev, "invalid TP_PARA_REG3(0x%x)\n", r); rc = EINVAL; } } v = V_HPZ0(0) | V_HPZ1(2) | V_HPZ2(4) | V_HPZ3(6); r = t4_read_reg(sc, A_ULP_RX_TDDP_PSZ); if (r != v) { device_printf(sc->dev, "invalid ULP_RX_TDDP_PSZ(0x%x)\n", r); rc = EINVAL; } m = v = F_TDDPTAGTCB; r = t4_read_reg(sc, A_ULP_RX_CTL); if ((r & m) != v) { device_printf(sc->dev, "invalid ULP_RX_CTL(0x%x)\n", r); rc = EINVAL; } m = V_INDICATESIZE(M_INDICATESIZE) | F_REARMDDPOFFSET | F_RESETDDPOFFSET; v = V_INDICATESIZE(indsz) | F_REARMDDPOFFSET | F_RESETDDPOFFSET; r = t4_read_reg(sc, A_TP_PARA_REG5); if ((r & m) != v) { device_printf(sc->dev, "invalid TP_PARA_REG5(0x%x)\n", r); rc = EINVAL; } r = t4_read_reg(sc, A_SGE_CONM_CTRL); s->fl_starve_threshold = G_EGRTHRESHOLD(r) * 2 + 1; if (is_t4(sc)) s->fl_starve_threshold2 = s->fl_starve_threshold; else s->fl_starve_threshold2 = G_EGRTHRESHOLDPACKING(r) * 2 + 1; /* egress queues: log2 of # of doorbells per BAR2 page */ r = t4_read_reg(sc, A_SGE_EGRESS_QUEUES_PER_PAGE_PF); r >>= S_QUEUESPERPAGEPF0 + (S_QUEUESPERPAGEPF1 - S_QUEUESPERPAGEPF0) * sc->pf; s->eq_s_qpp = r & M_QUEUESPERPAGEPF0; /* ingress queues: log2 of # of doorbells per BAR2 page */ r = t4_read_reg(sc, A_SGE_INGRESS_QUEUES_PER_PAGE_PF); r >>= S_QUEUESPERPAGEPF0 + (S_QUEUESPERPAGEPF1 - S_QUEUESPERPAGEPF0) * sc->pf; s->iq_s_qpp = r & M_QUEUESPERPAGEPF0; t4_init_tp_params(sc); t4_read_mtu_tbl(sc, sc->params.mtus, NULL); t4_load_mtus(sc, sc->params.mtus, sc->params.a_wnd, sc->params.b_wnd); return (rc); } int t4_create_dma_tag(struct adapter *sc) { int rc; rc = bus_dma_tag_create(bus_get_dma_tag(sc->dev), 1, 0, BUS_SPACE_MAXADDR, BUS_SPACE_MAXADDR, NULL, NULL, BUS_SPACE_MAXSIZE, BUS_SPACE_UNRESTRICTED, BUS_SPACE_MAXSIZE, BUS_DMA_ALLOCNOW, NULL, NULL, &sc->dmat); if (rc != 0) { device_printf(sc->dev, "failed to create main DMA tag: %d\n", rc); } return (rc); } void t4_sge_sysctls(struct adapter *sc, struct sysctl_ctx_list *ctx, struct sysctl_oid_list *children) { SYSCTL_ADD_PROC(ctx, children, OID_AUTO, "buffer_sizes", CTLTYPE_STRING | CTLFLAG_RD, &sc->sge, 0, sysctl_bufsizes, "A", "freelist buffer sizes"); SYSCTL_ADD_INT(ctx, children, OID_AUTO, "fl_pktshift", CTLFLAG_RD, NULL, fl_pktshift, "payload DMA offset in rx buffer (bytes)"); SYSCTL_ADD_INT(ctx, children, OID_AUTO, "fl_pad", CTLFLAG_RD, NULL, sc->sge.pad_boundary, "payload pad boundary (bytes)"); SYSCTL_ADD_INT(ctx, children, OID_AUTO, "spg_len", CTLFLAG_RD, NULL, spg_len, "status page size (bytes)"); SYSCTL_ADD_INT(ctx, children, OID_AUTO, "cong_drop", CTLFLAG_RD, NULL, cong_drop, "congestion drop setting"); SYSCTL_ADD_INT(ctx, children, OID_AUTO, "fl_pack", CTLFLAG_RD, NULL, sc->sge.pack_boundary, "payload pack boundary (bytes)"); } int t4_destroy_dma_tag(struct adapter *sc) { if (sc->dmat) bus_dma_tag_destroy(sc->dmat); return (0); } /* * Allocate and initialize the firmware event queue and the management queue. * * Returns errno on failure. Resources allocated up to that point may still be * allocated. Caller is responsible for cleanup in case this function fails. */ int t4_setup_adapter_queues(struct adapter *sc) { int rc; ADAPTER_LOCK_ASSERT_NOTOWNED(sc); sysctl_ctx_init(&sc->ctx); sc->flags |= ADAP_SYSCTL_CTX; /* * Firmware event queue */ rc = alloc_fwq(sc); if (rc != 0) return (rc); /* * Management queue. This is just a control queue that uses the fwq as * its associated iq. */ rc = alloc_mgmtq(sc); return (rc); } /* * Idempotent */ int t4_teardown_adapter_queues(struct adapter *sc) { ADAPTER_LOCK_ASSERT_NOTOWNED(sc); /* Do this before freeing the queue */ if (sc->flags & ADAP_SYSCTL_CTX) { sysctl_ctx_free(&sc->ctx); sc->flags &= ~ADAP_SYSCTL_CTX; } free_mgmtq(sc); free_fwq(sc); return (0); } static inline int port_intr_count(struct port_info *pi) { int rc = 0; if (pi->flags & INTR_RXQ) rc += pi->nrxq; #ifdef TCP_OFFLOAD if (pi->flags & INTR_OFLD_RXQ) rc += pi->nofldrxq; #endif #ifdef DEV_NETMAP if (pi->flags & INTR_NM_RXQ) rc += pi->nnmrxq; #endif return (rc); } static inline int first_vector(struct port_info *pi) { struct adapter *sc = pi->adapter; int rc = T4_EXTRA_INTR, i; if (sc->intr_count == 1) return (0); for_each_port(sc, i) { if (i == pi->port_id) break; rc += port_intr_count(sc->port[i]); } return (rc); } /* * Given an arbitrary "index," come up with an iq that can be used by other * queues (of this port) for interrupt forwarding, SGE egress updates, etc. * The iq returned is guaranteed to be something that takes direct interrupts. */ static struct sge_iq * port_intr_iq(struct port_info *pi, int idx) { struct adapter *sc = pi->adapter; struct sge *s = &sc->sge; struct sge_iq *iq = NULL; int nintr, i; if (sc->intr_count == 1) return (&sc->sge.fwq); nintr = port_intr_count(pi); KASSERT(nintr != 0, ("%s: pi %p has no exclusive interrupts, total interrupts = %d", __func__, pi, sc->intr_count)); #ifdef DEV_NETMAP /* Exclude netmap queues as they can't take anyone else's interrupts */ if (pi->flags & INTR_NM_RXQ) nintr -= pi->nnmrxq; KASSERT(nintr > 0, ("%s: pi %p has nintr %d after netmap adjustment of %d", __func__, pi, nintr, pi->nnmrxq)); #endif i = idx % nintr; if (pi->flags & INTR_RXQ) { if (i < pi->nrxq) { iq = &s->rxq[pi->first_rxq + i].iq; goto done; } i -= pi->nrxq; } #ifdef TCP_OFFLOAD if (pi->flags & INTR_OFLD_RXQ) { if (i < pi->nofldrxq) { iq = &s->ofld_rxq[pi->first_ofld_rxq + i].iq; goto done; } i -= pi->nofldrxq; } #endif panic("%s: pi %p, intr_flags 0x%lx, idx %d, total intr %d\n", __func__, pi, pi->flags & INTR_ALL, idx, nintr); done: MPASS(iq != NULL); KASSERT(iq->flags & IQ_INTR, ("%s: iq %p (port %p, intr_flags 0x%lx, idx %d)", __func__, iq, pi, pi->flags & INTR_ALL, idx)); return (iq); } /* Maximum payload that can be delivered with a single iq descriptor */ static inline int mtu_to_max_payload(struct adapter *sc, int mtu, const int toe) { int payload; #ifdef TCP_OFFLOAD if (toe) { payload = sc->tt.rx_coalesce ? G_RXCOALESCESIZE(t4_read_reg(sc, A_TP_PARA_REG2)) : mtu; } else { #endif /* large enough even when hw VLAN extraction is disabled */ payload = fl_pktshift + ETHER_HDR_LEN + ETHER_VLAN_ENCAP_LEN + mtu; #ifdef TCP_OFFLOAD } #endif return (payload); } int t4_setup_port_queues(struct port_info *pi) { int rc = 0, i, j, intr_idx, iqid; struct sge_rxq *rxq; struct sge_txq *txq; struct sge_wrq *ctrlq; #ifdef TCP_OFFLOAD struct sge_ofld_rxq *ofld_rxq; struct sge_wrq *ofld_txq; #endif #ifdef DEV_NETMAP struct sge_nm_rxq *nm_rxq; struct sge_nm_txq *nm_txq; #endif char name[16]; struct adapter *sc = pi->adapter; struct ifnet *ifp = pi->ifp; struct sysctl_oid *oid = device_get_sysctl_tree(pi->dev); struct sysctl_oid_list *children = SYSCTL_CHILDREN(oid); int maxp, mtu = ifp->if_mtu; /* Interrupt vector to start from (when using multiple vectors) */ intr_idx = first_vector(pi); /* * First pass over all NIC and TOE rx queues: * a) initialize iq and fl * b) allocate queue iff it will take direct interrupts. */ maxp = mtu_to_max_payload(sc, mtu, 0); if (pi->flags & INTR_RXQ) { oid = SYSCTL_ADD_NODE(&pi->ctx, children, OID_AUTO, "rxq", CTLFLAG_RD, NULL, "rx queues"); } for_each_rxq(pi, i, rxq) { init_iq(&rxq->iq, sc, pi->tmr_idx, pi->pktc_idx, pi->qsize_rxq); snprintf(name, sizeof(name), "%s rxq%d-fl", device_get_nameunit(pi->dev), i); init_fl(sc, &rxq->fl, pi->qsize_rxq / 8, maxp, name); if (pi->flags & INTR_RXQ) { rxq->iq.flags |= IQ_INTR; rc = alloc_rxq(pi, rxq, intr_idx, i, oid); if (rc != 0) goto done; intr_idx++; } } #ifdef TCP_OFFLOAD maxp = mtu_to_max_payload(sc, mtu, 1); if (is_offload(sc) && pi->flags & INTR_OFLD_RXQ) { oid = SYSCTL_ADD_NODE(&pi->ctx, children, OID_AUTO, "ofld_rxq", CTLFLAG_RD, NULL, "rx queues for offloaded TCP connections"); } for_each_ofld_rxq(pi, i, ofld_rxq) { init_iq(&ofld_rxq->iq, sc, pi->tmr_idx, pi->pktc_idx, pi->qsize_rxq); snprintf(name, sizeof(name), "%s ofld_rxq%d-fl", device_get_nameunit(pi->dev), i); init_fl(sc, &ofld_rxq->fl, pi->qsize_rxq / 8, maxp, name); if (pi->flags & INTR_OFLD_RXQ) { ofld_rxq->iq.flags |= IQ_INTR; rc = alloc_ofld_rxq(pi, ofld_rxq, intr_idx, i, oid); if (rc != 0) goto done; intr_idx++; } } #endif #ifdef DEV_NETMAP /* * We don't have buffers to back the netmap rx queues right now so we * create the queues in a way that doesn't set off any congestion signal * in the chip. */ if (pi->flags & INTR_NM_RXQ) { oid = SYSCTL_ADD_NODE(&pi->ctx, children, OID_AUTO, "nm_rxq", CTLFLAG_RD, NULL, "rx queues for netmap"); for_each_nm_rxq(pi, i, nm_rxq) { rc = alloc_nm_rxq(pi, nm_rxq, intr_idx, i, oid); if (rc != 0) goto done; intr_idx++; } } #endif /* * Second pass over all NIC and TOE rx queues. The queues forwarding * their interrupts are allocated now. */ j = 0; if (!(pi->flags & INTR_RXQ)) { oid = SYSCTL_ADD_NODE(&pi->ctx, children, OID_AUTO, "rxq", CTLFLAG_RD, NULL, "rx queues"); for_each_rxq(pi, i, rxq) { MPASS(!(rxq->iq.flags & IQ_INTR)); intr_idx = port_intr_iq(pi, j)->abs_id; rc = alloc_rxq(pi, rxq, intr_idx, i, oid); if (rc != 0) goto done; j++; } } #ifdef TCP_OFFLOAD if (is_offload(sc) && !(pi->flags & INTR_OFLD_RXQ)) { oid = SYSCTL_ADD_NODE(&pi->ctx, children, OID_AUTO, "ofld_rxq", CTLFLAG_RD, NULL, "rx queues for offloaded TCP connections"); for_each_ofld_rxq(pi, i, ofld_rxq) { MPASS(!(ofld_rxq->iq.flags & IQ_INTR)); intr_idx = port_intr_iq(pi, j)->abs_id; rc = alloc_ofld_rxq(pi, ofld_rxq, intr_idx, i, oid); if (rc != 0) goto done; j++; } } #endif #ifdef DEV_NETMAP if (!(pi->flags & INTR_NM_RXQ)) CXGBE_UNIMPLEMENTED(__func__); #endif /* * Now the tx queues. Only one pass needed. */ oid = SYSCTL_ADD_NODE(&pi->ctx, children, OID_AUTO, "txq", CTLFLAG_RD, NULL, "tx queues"); j = 0; for_each_txq(pi, i, txq) { iqid = port_intr_iq(pi, j)->cntxt_id; snprintf(name, sizeof(name), "%s txq%d", device_get_nameunit(pi->dev), i); init_eq(&txq->eq, EQ_ETH, pi->qsize_txq, pi->tx_chan, iqid, name); rc = alloc_txq(pi, txq, i, oid); if (rc != 0) goto done; j++; } #ifdef TCP_OFFLOAD oid = SYSCTL_ADD_NODE(&pi->ctx, children, OID_AUTO, "ofld_txq", CTLFLAG_RD, NULL, "tx queues for offloaded TCP connections"); for_each_ofld_txq(pi, i, ofld_txq) { struct sysctl_oid *oid2; iqid = port_intr_iq(pi, j)->cntxt_id; snprintf(name, sizeof(name), "%s ofld_txq%d", device_get_nameunit(pi->dev), i); init_eq(&ofld_txq->eq, EQ_OFLD, pi->qsize_txq, pi->tx_chan, iqid, name); snprintf(name, sizeof(name), "%d", i); oid2 = SYSCTL_ADD_NODE(&pi->ctx, SYSCTL_CHILDREN(oid), OID_AUTO, name, CTLFLAG_RD, NULL, "offload tx queue"); rc = alloc_wrq(sc, pi, ofld_txq, oid2); if (rc != 0) goto done; j++; } #endif #ifdef DEV_NETMAP oid = SYSCTL_ADD_NODE(&pi->ctx, children, OID_AUTO, "nm_txq", CTLFLAG_RD, NULL, "tx queues for netmap use"); for_each_nm_txq(pi, i, nm_txq) { iqid = pi->first_nm_rxq + (j % pi->nnmrxq); rc = alloc_nm_txq(pi, nm_txq, iqid, i, oid); if (rc != 0) goto done; j++; } #endif /* * Finally, the control queue. */ oid = SYSCTL_ADD_NODE(&pi->ctx, children, OID_AUTO, "ctrlq", CTLFLAG_RD, NULL, "ctrl queue"); ctrlq = &sc->sge.ctrlq[pi->port_id]; iqid = port_intr_iq(pi, 0)->cntxt_id; snprintf(name, sizeof(name), "%s ctrlq", device_get_nameunit(pi->dev)); init_eq(&ctrlq->eq, EQ_CTRL, CTRL_EQ_QSIZE, pi->tx_chan, iqid, name); rc = alloc_wrq(sc, pi, ctrlq, oid); done: if (rc) t4_teardown_port_queues(pi); return (rc); } /* * Idempotent */ int t4_teardown_port_queues(struct port_info *pi) { int i; struct adapter *sc = pi->adapter; struct sge_rxq *rxq; struct sge_txq *txq; #ifdef TCP_OFFLOAD struct sge_ofld_rxq *ofld_rxq; struct sge_wrq *ofld_txq; #endif #ifdef DEV_NETMAP struct sge_nm_rxq *nm_rxq; struct sge_nm_txq *nm_txq; #endif /* Do this before freeing the queues */ if (pi->flags & PORT_SYSCTL_CTX) { sysctl_ctx_free(&pi->ctx); pi->flags &= ~PORT_SYSCTL_CTX; } /* * Take down all the tx queues first, as they reference the rx queues * (for egress updates, etc.). */ free_wrq(sc, &sc->sge.ctrlq[pi->port_id]); for_each_txq(pi, i, txq) { free_txq(pi, txq); } #ifdef TCP_OFFLOAD for_each_ofld_txq(pi, i, ofld_txq) { free_wrq(sc, ofld_txq); } #endif #ifdef DEV_NETMAP for_each_nm_txq(pi, i, nm_txq) free_nm_txq(pi, nm_txq); #endif /* * Then take down the rx queues that forward their interrupts, as they * reference other rx queues. */ for_each_rxq(pi, i, rxq) { if ((rxq->iq.flags & IQ_INTR) == 0) free_rxq(pi, rxq); } #ifdef TCP_OFFLOAD for_each_ofld_rxq(pi, i, ofld_rxq) { if ((ofld_rxq->iq.flags & IQ_INTR) == 0) free_ofld_rxq(pi, ofld_rxq); } #endif #ifdef DEV_NETMAP for_each_nm_rxq(pi, i, nm_rxq) free_nm_rxq(pi, nm_rxq); #endif /* * Then take down the rx queues that take direct interrupts. */ for_each_rxq(pi, i, rxq) { if (rxq->iq.flags & IQ_INTR) free_rxq(pi, rxq); } #ifdef TCP_OFFLOAD for_each_ofld_rxq(pi, i, ofld_rxq) { if (ofld_rxq->iq.flags & IQ_INTR) free_ofld_rxq(pi, ofld_rxq); } #endif return (0); } /* * Deals with errors and the firmware event queue. All data rx queues forward * their interrupt to the firmware event queue. */ void t4_intr_all(void *arg) { struct adapter *sc = arg; struct sge_iq *fwq = &sc->sge.fwq; t4_intr_err(arg); if (atomic_cmpset_int(&fwq->state, IQS_IDLE, IQS_BUSY)) { service_iq(fwq, 0); atomic_cmpset_int(&fwq->state, IQS_BUSY, IQS_IDLE); } } /* Deals with error interrupts */ void t4_intr_err(void *arg) { struct adapter *sc = arg; t4_write_reg(sc, MYPF_REG(A_PCIE_PF_CLI), 0); t4_slow_intr_handler(sc); } void t4_intr_evt(void *arg) { struct sge_iq *iq = arg; if (atomic_cmpset_int(&iq->state, IQS_IDLE, IQS_BUSY)) { service_iq(iq, 0); atomic_cmpset_int(&iq->state, IQS_BUSY, IQS_IDLE); } } void t4_intr(void *arg) { struct sge_iq *iq = arg; if (atomic_cmpset_int(&iq->state, IQS_IDLE, IQS_BUSY)) { service_iq(iq, 0); atomic_cmpset_int(&iq->state, IQS_BUSY, IQS_IDLE); } } /* * Deals with anything and everything on the given ingress queue. */ static int service_iq(struct sge_iq *iq, int budget) { struct sge_iq *q; struct sge_rxq *rxq = iq_to_rxq(iq); /* Use iff iq is part of rxq */ struct sge_fl *fl; /* Use iff IQ_HAS_FL */ struct adapter *sc = iq->adapter; struct iq_desc *d = &iq->desc[iq->cidx]; int ndescs = 0, limit; int rsp_type, refill; uint32_t lq; uint16_t fl_hw_cidx; struct mbuf *m0; STAILQ_HEAD(, sge_iq) iql = STAILQ_HEAD_INITIALIZER(iql); #if defined(INET) || defined(INET6) const struct timeval lro_timeout = {0, sc->lro_timeout}; #endif KASSERT(iq->state == IQS_BUSY, ("%s: iq %p not BUSY", __func__, iq)); limit = budget ? budget : iq->qsize / 16; if (iq->flags & IQ_HAS_FL) { fl = &rxq->fl; fl_hw_cidx = fl->hw_cidx; /* stable snapshot */ } else { fl = NULL; fl_hw_cidx = 0; /* to silence gcc warning */ } /* * We always come back and check the descriptor ring for new indirect * interrupts and other responses after running a single handler. */ for (;;) { while ((d->rsp.u.type_gen & F_RSPD_GEN) == iq->gen) { rmb(); refill = 0; m0 = NULL; rsp_type = G_RSPD_TYPE(d->rsp.u.type_gen); lq = be32toh(d->rsp.pldbuflen_qid); switch (rsp_type) { case X_RSPD_TYPE_FLBUF: KASSERT(iq->flags & IQ_HAS_FL, ("%s: data for an iq (%p) with no freelist", __func__, iq)); m0 = get_fl_payload(sc, fl, lq); if (__predict_false(m0 == NULL)) goto process_iql; refill = IDXDIFF(fl->hw_cidx, fl_hw_cidx, fl->sidx) > 2; #ifdef T4_PKT_TIMESTAMP /* * 60 bit timestamp for the payload is * *(uint64_t *)m0->m_pktdat. Note that it is * in the leading free-space in the mbuf. The * kernel can clobber it during a pullup, * m_copymdata, etc. You need to make sure that * the mbuf reaches you unmolested if you care * about the timestamp. */ *(uint64_t *)m0->m_pktdat = be64toh(ctrl->u.last_flit) & 0xfffffffffffffff; #endif /* fall through */ case X_RSPD_TYPE_CPL: KASSERT(d->rss.opcode < NUM_CPL_CMDS, ("%s: bad opcode %02x.", __func__, d->rss.opcode)); sc->cpl_handler[d->rss.opcode](iq, &d->rss, m0); break; case X_RSPD_TYPE_INTR: /* * Interrupts should be forwarded only to queues * that are not forwarding their interrupts. * This means service_iq can recurse but only 1 * level deep. */ KASSERT(budget == 0, ("%s: budget %u, rsp_type %u", __func__, budget, rsp_type)); /* * There are 1K interrupt-capable queues (qids 0 * through 1023). A response type indicating a * forwarded interrupt with a qid >= 1K is an * iWARP async notification. */ if (lq >= 1024) { sc->an_handler(iq, &d->rsp); break; } q = sc->sge.iqmap[lq - sc->sge.iq_start]; if (atomic_cmpset_int(&q->state, IQS_IDLE, IQS_BUSY)) { if (service_iq(q, q->qsize / 16) == 0) { atomic_cmpset_int(&q->state, IQS_BUSY, IQS_IDLE); } else { STAILQ_INSERT_TAIL(&iql, q, link); } } break; default: KASSERT(0, ("%s: illegal response type %d on iq %p", __func__, rsp_type, iq)); log(LOG_ERR, "%s: illegal response type %d on iq %p", device_get_nameunit(sc->dev), rsp_type, iq); break; } d++; if (__predict_false(++iq->cidx == iq->sidx)) { iq->cidx = 0; iq->gen ^= F_RSPD_GEN; d = &iq->desc[0]; } if (__predict_false(++ndescs == limit)) { t4_write_reg(sc, MYPF_REG(A_SGE_PF_GTS), V_CIDXINC(ndescs) | V_INGRESSQID(iq->cntxt_id) | V_SEINTARM(V_QINTR_TIMER_IDX(X_TIMERREG_UPDATE_CIDX))); ndescs = 0; #if defined(INET) || defined(INET6) if (iq->flags & IQ_LRO_ENABLED && sc->lro_timeout != 0) { tcp_lro_flush_inactive(&rxq->lro, &lro_timeout); } #endif if (budget) { if (iq->flags & IQ_HAS_FL) { FL_LOCK(fl); refill_fl(sc, fl, 32); FL_UNLOCK(fl); } return (EINPROGRESS); } } if (refill) { FL_LOCK(fl); refill_fl(sc, fl, 32); FL_UNLOCK(fl); fl_hw_cidx = fl->hw_cidx; } } process_iql: if (STAILQ_EMPTY(&iql)) break; /* * Process the head only, and send it to the back of the list if * it's still not done. */ q = STAILQ_FIRST(&iql); STAILQ_REMOVE_HEAD(&iql, link); if (service_iq(q, q->qsize / 8) == 0) atomic_cmpset_int(&q->state, IQS_BUSY, IQS_IDLE); else STAILQ_INSERT_TAIL(&iql, q, link); } #if defined(INET) || defined(INET6) if (iq->flags & IQ_LRO_ENABLED) { struct lro_ctrl *lro = &rxq->lro; struct lro_entry *l; while (!SLIST_EMPTY(&lro->lro_active)) { l = SLIST_FIRST(&lro->lro_active); SLIST_REMOVE_HEAD(&lro->lro_active, next); tcp_lro_flush(lro, l); } } #endif t4_write_reg(sc, MYPF_REG(A_SGE_PF_GTS), V_CIDXINC(ndescs) | V_INGRESSQID((u32)iq->cntxt_id) | V_SEINTARM(iq->intr_params)); if (iq->flags & IQ_HAS_FL) { int starved; FL_LOCK(fl); starved = refill_fl(sc, fl, 64); FL_UNLOCK(fl); if (__predict_false(starved != 0)) add_fl_to_sfl(sc, fl); } return (0); } static inline int cl_has_metadata(struct sge_fl *fl, struct cluster_layout *cll) { int rc = fl->flags & FL_BUF_PACKING || cll->region1 > 0; if (rc) MPASS(cll->region3 >= CL_METADATA_SIZE); return (rc); } static inline struct cluster_metadata * cl_metadata(struct adapter *sc, struct sge_fl *fl, struct cluster_layout *cll, caddr_t cl) { if (cl_has_metadata(fl, cll)) { struct sw_zone_info *swz = &sc->sge.sw_zone_info[cll->zidx]; return ((struct cluster_metadata *)(cl + swz->size) - 1); } return (NULL); } static void rxb_free(struct mbuf *m, void *arg1, void *arg2) { uma_zone_t zone = arg1; caddr_t cl = arg2; uma_zfree(zone, cl); counter_u64_add(extfree_rels, 1); } /* * The mbuf returned by this function could be allocated from zone_mbuf or * constructed in spare room in the cluster. * * The mbuf carries the payload in one of these ways * a) frame inside the mbuf (mbuf from zone_mbuf) * b) m_cljset (for clusters without metadata) zone_mbuf * c) m_extaddref (cluster with metadata) inline mbuf * d) m_extaddref (cluster with metadata) zone_mbuf */ static struct mbuf * get_scatter_segment(struct adapter *sc, struct sge_fl *fl, int fr_offset, int remaining) { struct mbuf *m; struct fl_sdesc *sd = &fl->sdesc[fl->cidx]; struct cluster_layout *cll = &sd->cll; struct sw_zone_info *swz = &sc->sge.sw_zone_info[cll->zidx]; struct hw_buf_info *hwb = &sc->sge.hw_buf_info[cll->hwidx]; struct cluster_metadata *clm = cl_metadata(sc, fl, cll, sd->cl); int len, blen; caddr_t payload; blen = hwb->size - fl->rx_offset; /* max possible in this buf */ len = min(remaining, blen); payload = sd->cl + cll->region1 + fl->rx_offset; if (fl->flags & FL_BUF_PACKING) { const u_int l = fr_offset + len; const u_int pad = roundup2(l, fl->buf_boundary) - l; if (fl->rx_offset + len + pad < hwb->size) blen = len + pad; MPASS(fl->rx_offset + blen <= hwb->size); } else { MPASS(fl->rx_offset == 0); /* not packing */ } if (sc->sc_do_rxcopy && len < RX_COPY_THRESHOLD) { /* * Copy payload into a freshly allocated mbuf. */ m = fr_offset == 0 ? m_gethdr(M_NOWAIT, MT_DATA) : m_get(M_NOWAIT, MT_DATA); if (m == NULL) return (NULL); fl->mbuf_allocated++; #ifdef T4_PKT_TIMESTAMP /* Leave room for a timestamp */ m->m_data += 8; #endif /* copy data to mbuf */ bcopy(payload, mtod(m, caddr_t), len); } else if (sd->nmbuf * MSIZE < cll->region1) { /* * There's spare room in the cluster for an mbuf. Create one * and associate it with the payload that's in the cluster. */ MPASS(clm != NULL); m = (struct mbuf *)(sd->cl + sd->nmbuf * MSIZE); /* No bzero required */ if (m_init(m, NULL, 0, M_NOWAIT, MT_DATA, fr_offset == 0 ? M_PKTHDR | M_NOFREE : M_NOFREE)) return (NULL); fl->mbuf_inlined++; m_extaddref(m, payload, blen, &clm->refcount, rxb_free, swz->zone, sd->cl); if (sd->nmbuf++ == 0) counter_u64_add(extfree_refs, 1); } else { /* * Grab an mbuf from zone_mbuf and associate it with the * payload in the cluster. */ m = fr_offset == 0 ? m_gethdr(M_NOWAIT, MT_DATA) : m_get(M_NOWAIT, MT_DATA); if (m == NULL) return (NULL); fl->mbuf_allocated++; if (clm != NULL) { m_extaddref(m, payload, blen, &clm->refcount, rxb_free, swz->zone, sd->cl); if (sd->nmbuf++ == 0) counter_u64_add(extfree_refs, 1); } else { m_cljset(m, sd->cl, swz->type); sd->cl = NULL; /* consumed, not a recycle candidate */ } } if (fr_offset == 0) m->m_pkthdr.len = remaining; m->m_len = len; if (fl->flags & FL_BUF_PACKING) { fl->rx_offset += blen; MPASS(fl->rx_offset <= hwb->size); if (fl->rx_offset < hwb->size) return (m); /* without advancing the cidx */ } if (__predict_false(++fl->cidx % 8 == 0)) { uint16_t cidx = fl->cidx / 8; if (__predict_false(cidx == fl->sidx)) fl->cidx = cidx = 0; fl->hw_cidx = cidx; } fl->rx_offset = 0; return (m); } static struct mbuf * get_fl_payload(struct adapter *sc, struct sge_fl *fl, uint32_t len_newbuf) { struct mbuf *m0, *m, **pnext; u_int remaining; const u_int total = G_RSPD_LEN(len_newbuf); if (__predict_false(fl->flags & FL_BUF_RESUME)) { M_ASSERTPKTHDR(fl->m0); MPASS(fl->m0->m_pkthdr.len == total); MPASS(fl->remaining < total); m0 = fl->m0; pnext = fl->pnext; remaining = fl->remaining; fl->flags &= ~FL_BUF_RESUME; goto get_segment; } if (fl->rx_offset > 0 && len_newbuf & F_RSPD_NEWBUF) { fl->rx_offset = 0; if (__predict_false(++fl->cidx % 8 == 0)) { uint16_t cidx = fl->cidx / 8; if (__predict_false(cidx == fl->sidx)) fl->cidx = cidx = 0; fl->hw_cidx = cidx; } } /* * Payload starts at rx_offset in the current hw buffer. Its length is * 'len' and it may span multiple hw buffers. */ m0 = get_scatter_segment(sc, fl, 0, total); if (m0 == NULL) return (NULL); remaining = total - m0->m_len; pnext = &m0->m_next; while (remaining > 0) { get_segment: MPASS(fl->rx_offset == 0); m = get_scatter_segment(sc, fl, total - remaining, remaining); if (__predict_false(m == NULL)) { fl->m0 = m0; fl->pnext = pnext; fl->remaining = remaining; fl->flags |= FL_BUF_RESUME; return (NULL); } *pnext = m; pnext = &m->m_next; remaining -= m->m_len; } *pnext = NULL; return (m0); } static int t4_eth_rx(struct sge_iq *iq, const struct rss_header *rss, struct mbuf *m0) { struct sge_rxq *rxq = iq_to_rxq(iq); struct ifnet *ifp = rxq->ifp; const struct cpl_rx_pkt *cpl = (const void *)(rss + 1); #if defined(INET) || defined(INET6) struct lro_ctrl *lro = &rxq->lro; #endif static const int sw_hashtype[4][2] = { {M_HASHTYPE_NONE, M_HASHTYPE_NONE}, {M_HASHTYPE_RSS_IPV4, M_HASHTYPE_RSS_IPV6}, {M_HASHTYPE_RSS_TCP_IPV4, M_HASHTYPE_RSS_TCP_IPV6}, {M_HASHTYPE_RSS_UDP_IPV4, M_HASHTYPE_RSS_UDP_IPV6}, }; KASSERT(m0 != NULL, ("%s: no payload with opcode %02x", __func__, rss->opcode)); m0->m_pkthdr.len -= fl_pktshift; m0->m_len -= fl_pktshift; m0->m_data += fl_pktshift; m0->m_pkthdr.rcvif = ifp; M_HASHTYPE_SET(m0, sw_hashtype[rss->hash_type][rss->ipv6]); m0->m_pkthdr.flowid = be32toh(rss->hash_val); if (cpl->csum_calc && !cpl->err_vec) { if (ifp->if_capenable & IFCAP_RXCSUM && cpl->l2info & htobe32(F_RXF_IP)) { m0->m_pkthdr.csum_flags = (CSUM_IP_CHECKED | CSUM_IP_VALID | CSUM_DATA_VALID | CSUM_PSEUDO_HDR); rxq->rxcsum++; } else if (ifp->if_capenable & IFCAP_RXCSUM_IPV6 && cpl->l2info & htobe32(F_RXF_IP6)) { m0->m_pkthdr.csum_flags = (CSUM_DATA_VALID_IPV6 | CSUM_PSEUDO_HDR); rxq->rxcsum++; } if (__predict_false(cpl->ip_frag)) m0->m_pkthdr.csum_data = be16toh(cpl->csum); else m0->m_pkthdr.csum_data = 0xffff; } if (cpl->vlan_ex) { m0->m_pkthdr.ether_vtag = be16toh(cpl->vlan); m0->m_flags |= M_VLANTAG; rxq->vlan_extraction++; } #if defined(INET) || defined(INET6) if (cpl->l2info & htobe32(F_RXF_LRO) && iq->flags & IQ_LRO_ENABLED && tcp_lro_rx(lro, m0, 0) == 0) { /* queued for LRO */ } else #endif ifp->if_input(ifp, m0); return (0); } /* * Must drain the wrq or make sure that someone else will. */ static void wrq_tx_drain(void *arg, int n) { struct sge_wrq *wrq = arg; struct sge_eq *eq = &wrq->eq; EQ_LOCK(eq); if (TAILQ_EMPTY(&wrq->incomplete_wrs) && !STAILQ_EMPTY(&wrq->wr_list)) drain_wrq_wr_list(wrq->adapter, wrq); EQ_UNLOCK(eq); } static void drain_wrq_wr_list(struct adapter *sc, struct sge_wrq *wrq) { struct sge_eq *eq = &wrq->eq; u_int available, dbdiff; /* # of hardware descriptors */ u_int n; struct wrqe *wr; struct fw_eth_tx_pkt_wr *dst; /* any fw WR struct will do */ EQ_LOCK_ASSERT_OWNED(eq); MPASS(TAILQ_EMPTY(&wrq->incomplete_wrs)); wr = STAILQ_FIRST(&wrq->wr_list); MPASS(wr != NULL); /* Must be called with something useful to do */ dbdiff = IDXDIFF(eq->pidx, eq->dbidx, eq->sidx); do { eq->cidx = read_hw_cidx(eq); if (eq->pidx == eq->cidx) available = eq->sidx - 1; else available = IDXDIFF(eq->cidx, eq->pidx, eq->sidx) - 1; MPASS(wr->wrq == wrq); n = howmany(wr->wr_len, EQ_ESIZE); if (available < n) return; dst = (void *)&eq->desc[eq->pidx]; if (__predict_true(eq->sidx - eq->pidx > n)) { /* Won't wrap, won't end exactly at the status page. */ bcopy(&wr->wr[0], dst, wr->wr_len); eq->pidx += n; } else { int first_portion = (eq->sidx - eq->pidx) * EQ_ESIZE; bcopy(&wr->wr[0], dst, first_portion); if (wr->wr_len > first_portion) { bcopy(&wr->wr[first_portion], &eq->desc[0], wr->wr_len - first_portion); } eq->pidx = n - (eq->sidx - eq->pidx); } if (available < eq->sidx / 4 && atomic_cmpset_int(&eq->equiq, 0, 1)) { dst->equiq_to_len16 |= htobe32(F_FW_WR_EQUIQ | F_FW_WR_EQUEQ); eq->equeqidx = eq->pidx; } else if (IDXDIFF(eq->pidx, eq->equeqidx, eq->sidx) >= 32) { dst->equiq_to_len16 |= htobe32(F_FW_WR_EQUEQ); eq->equeqidx = eq->pidx; } dbdiff += n; if (dbdiff >= 16) { ring_eq_db(sc, eq, dbdiff); dbdiff = 0; } STAILQ_REMOVE_HEAD(&wrq->wr_list, link); free_wrqe(wr); MPASS(wrq->nwr_pending > 0); wrq->nwr_pending--; MPASS(wrq->ndesc_needed >= n); wrq->ndesc_needed -= n; } while ((wr = STAILQ_FIRST(&wrq->wr_list)) != NULL); if (dbdiff) ring_eq_db(sc, eq, dbdiff); } /* * Doesn't fail. Holds on to work requests it can't send right away. */ void t4_wrq_tx_locked(struct adapter *sc, struct sge_wrq *wrq, struct wrqe *wr) { #ifdef INVARIANTS struct sge_eq *eq = &wrq->eq; #endif EQ_LOCK_ASSERT_OWNED(eq); MPASS(wr != NULL); MPASS(wr->wr_len > 0 && wr->wr_len <= SGE_MAX_WR_LEN); MPASS((wr->wr_len & 0x7) == 0); STAILQ_INSERT_TAIL(&wrq->wr_list, wr, link); wrq->nwr_pending++; wrq->ndesc_needed += howmany(wr->wr_len, EQ_ESIZE); if (!TAILQ_EMPTY(&wrq->incomplete_wrs)) return; /* commit_wrq_wr will drain wr_list as well. */ drain_wrq_wr_list(sc, wrq); /* Doorbell must have caught up to the pidx. */ MPASS(eq->pidx == eq->dbidx); } void t4_update_fl_bufsize(struct ifnet *ifp) { struct port_info *pi = ifp->if_softc; struct adapter *sc = pi->adapter; struct sge_rxq *rxq; #ifdef TCP_OFFLOAD struct sge_ofld_rxq *ofld_rxq; #endif struct sge_fl *fl; int i, maxp, mtu = ifp->if_mtu; maxp = mtu_to_max_payload(sc, mtu, 0); for_each_rxq(pi, i, rxq) { fl = &rxq->fl; FL_LOCK(fl); find_best_refill_source(sc, fl, maxp); FL_UNLOCK(fl); } #ifdef TCP_OFFLOAD maxp = mtu_to_max_payload(sc, mtu, 1); for_each_ofld_rxq(pi, i, ofld_rxq) { fl = &ofld_rxq->fl; FL_LOCK(fl); find_best_refill_source(sc, fl, maxp); FL_UNLOCK(fl); } #endif } static inline int mbuf_nsegs(struct mbuf *m) { M_ASSERTPKTHDR(m); KASSERT(m->m_pkthdr.l5hlen > 0, ("%s: mbuf %p missing information on # of segments.", __func__, m)); return (m->m_pkthdr.l5hlen); } static inline void set_mbuf_nsegs(struct mbuf *m, uint8_t nsegs) { M_ASSERTPKTHDR(m); m->m_pkthdr.l5hlen = nsegs; } static inline int mbuf_len16(struct mbuf *m) { int n; M_ASSERTPKTHDR(m); n = m->m_pkthdr.PH_loc.eight[0]; MPASS(n > 0 && n <= SGE_MAX_WR_LEN / 16); return (n); } static inline void set_mbuf_len16(struct mbuf *m, uint8_t len16) { M_ASSERTPKTHDR(m); m->m_pkthdr.PH_loc.eight[0] = len16; } static inline int needs_tso(struct mbuf *m) { M_ASSERTPKTHDR(m); if (m->m_pkthdr.csum_flags & CSUM_TSO) { KASSERT(m->m_pkthdr.tso_segsz > 0, ("%s: TSO requested in mbuf %p but MSS not provided", __func__, m)); return (1); } return (0); } static inline int needs_l3_csum(struct mbuf *m) { M_ASSERTPKTHDR(m); if (m->m_pkthdr.csum_flags & (CSUM_IP | CSUM_TSO)) return (1); return (0); } static inline int needs_l4_csum(struct mbuf *m) { M_ASSERTPKTHDR(m); if (m->m_pkthdr.csum_flags & (CSUM_TCP | CSUM_UDP | CSUM_UDP_IPV6 | CSUM_TCP_IPV6 | CSUM_TSO)) return (1); return (0); } static inline int needs_vlan_insertion(struct mbuf *m) { M_ASSERTPKTHDR(m); if (m->m_flags & M_VLANTAG) { KASSERT(m->m_pkthdr.ether_vtag != 0, ("%s: HWVLAN requested in mbuf %p but tag not provided", __func__, m)); return (1); } return (0); } static void * m_advance(struct mbuf **pm, int *poffset, int len) { struct mbuf *m = *pm; int offset = *poffset; uintptr_t p = 0; MPASS(len > 0); while (len) { if (offset + len < m->m_len) { offset += len; p = mtod(m, uintptr_t) + offset; break; } len -= m->m_len - offset; m = m->m_next; offset = 0; MPASS(m != NULL); } *poffset = offset; *pm = m; return ((void *)p); } static inline int same_paddr(char *a, char *b) { if (a == b) return (1); else if (a != NULL && b != NULL) { vm_offset_t x = (vm_offset_t)a; vm_offset_t y = (vm_offset_t)b; if ((x & PAGE_MASK) == (y & PAGE_MASK) && pmap_kextract(x) == pmap_kextract(y)) return (1); } return (0); } /* * Can deal with empty mbufs in the chain that have m_len = 0, but the chain * must have at least one mbuf that's not empty. */ static inline int count_mbuf_nsegs(struct mbuf *m) { char *prev_end, *start; int len, nsegs; MPASS(m != NULL); nsegs = 0; prev_end = NULL; for (; m; m = m->m_next) { len = m->m_len; if (__predict_false(len == 0)) continue; start = mtod(m, char *); nsegs += sglist_count(start, len); if (same_paddr(prev_end, start)) nsegs--; prev_end = start + len; } MPASS(nsegs > 0); return (nsegs); } /* * Analyze the mbuf to determine its tx needs. The mbuf passed in may change: * a) caller can assume it's been freed if this function returns with an error. * b) it may get defragged up if the gather list is too long for the hardware. */ int parse_pkt(struct mbuf **mp) { struct mbuf *m0 = *mp, *m; int rc, nsegs, defragged = 0, offset; struct ether_header *eh; void *l3hdr; #if defined(INET) || defined(INET6) struct tcphdr *tcp; #endif uint16_t eh_type; M_ASSERTPKTHDR(m0); if (__predict_false(m0->m_pkthdr.len < ETHER_HDR_LEN)) { rc = EINVAL; fail: m_freem(m0); *mp = NULL; return (rc); } restart: /* * First count the number of gather list segments in the payload. * Defrag the mbuf if nsegs exceeds the hardware limit. */ M_ASSERTPKTHDR(m0); MPASS(m0->m_pkthdr.len > 0); nsegs = count_mbuf_nsegs(m0); if (nsegs > (needs_tso(m0) ? TX_SGL_SEGS_TSO : TX_SGL_SEGS)) { if (defragged++ > 0 || (m = m_defrag(m0, M_NOWAIT)) == NULL) { rc = EFBIG; goto fail; } *mp = m0 = m; /* update caller's copy after defrag */ goto restart; } if (__predict_false(nsegs > 2 && m0->m_pkthdr.len <= MHLEN)) { m0 = m_pullup(m0, m0->m_pkthdr.len); if (m0 == NULL) { /* Should have left well enough alone. */ rc = EFBIG; goto fail; } *mp = m0; /* update caller's copy after pullup */ goto restart; } set_mbuf_nsegs(m0, nsegs); set_mbuf_len16(m0, txpkt_len16(nsegs, needs_tso(m0))); if (!needs_tso(m0)) return (0); m = m0; eh = mtod(m, struct ether_header *); eh_type = ntohs(eh->ether_type); if (eh_type == ETHERTYPE_VLAN) { struct ether_vlan_header *evh = (void *)eh; eh_type = ntohs(evh->evl_proto); m0->m_pkthdr.l2hlen = sizeof(*evh); } else m0->m_pkthdr.l2hlen = sizeof(*eh); offset = 0; l3hdr = m_advance(&m, &offset, m0->m_pkthdr.l2hlen); switch (eh_type) { #ifdef INET6 case ETHERTYPE_IPV6: { struct ip6_hdr *ip6 = l3hdr; MPASS(ip6->ip6_nxt == IPPROTO_TCP); m0->m_pkthdr.l3hlen = sizeof(*ip6); break; } #endif #ifdef INET case ETHERTYPE_IP: { struct ip *ip = l3hdr; m0->m_pkthdr.l3hlen = ip->ip_hl * 4; break; } #endif default: panic("%s: ethertype 0x%04x unknown. if_cxgbe must be compiled" " with the same INET/INET6 options as the kernel.", __func__, eh_type); } #if defined(INET) || defined(INET6) tcp = m_advance(&m, &offset, m0->m_pkthdr.l3hlen); m0->m_pkthdr.l4hlen = tcp->th_off * 4; #endif MPASS(m0 == *mp); return (0); } void * start_wrq_wr(struct sge_wrq *wrq, int len16, struct wrq_cookie *cookie) { struct sge_eq *eq = &wrq->eq; struct adapter *sc = wrq->adapter; int ndesc, available; struct wrqe *wr; void *w; MPASS(len16 > 0); ndesc = howmany(len16, EQ_ESIZE / 16); MPASS(ndesc > 0 && ndesc <= SGE_MAX_WR_NDESC); EQ_LOCK(eq); if (!STAILQ_EMPTY(&wrq->wr_list)) drain_wrq_wr_list(sc, wrq); if (!STAILQ_EMPTY(&wrq->wr_list)) { slowpath: EQ_UNLOCK(eq); wr = alloc_wrqe(len16 * 16, wrq); if (__predict_false(wr == NULL)) return (NULL); cookie->pidx = -1; cookie->ndesc = ndesc; return (&wr->wr); } eq->cidx = read_hw_cidx(eq); if (eq->pidx == eq->cidx) available = eq->sidx - 1; else available = IDXDIFF(eq->cidx, eq->pidx, eq->sidx) - 1; if (available < ndesc) goto slowpath; cookie->pidx = eq->pidx; cookie->ndesc = ndesc; TAILQ_INSERT_TAIL(&wrq->incomplete_wrs, cookie, link); w = &eq->desc[eq->pidx]; IDXINCR(eq->pidx, ndesc, eq->sidx); if (__predict_false(eq->pidx < ndesc - 1)) { w = &wrq->ss[0]; wrq->ss_pidx = cookie->pidx; wrq->ss_len = len16 * 16; } EQ_UNLOCK(eq); return (w); } void commit_wrq_wr(struct sge_wrq *wrq, void *w, struct wrq_cookie *cookie) { struct sge_eq *eq = &wrq->eq; struct adapter *sc = wrq->adapter; int ndesc, pidx; struct wrq_cookie *prev, *next; if (cookie->pidx == -1) { struct wrqe *wr = __containerof(w, struct wrqe, wr); t4_wrq_tx(sc, wr); return; } ndesc = cookie->ndesc; /* Can be more than SGE_MAX_WR_NDESC here. */ pidx = cookie->pidx; MPASS(pidx >= 0 && pidx < eq->sidx); if (__predict_false(w == &wrq->ss[0])) { int n = (eq->sidx - wrq->ss_pidx) * EQ_ESIZE; MPASS(wrq->ss_len > n); /* WR had better wrap around. */ bcopy(&wrq->ss[0], &eq->desc[wrq->ss_pidx], n); bcopy(&wrq->ss[n], &eq->desc[0], wrq->ss_len - n); wrq->tx_wrs_ss++; } else wrq->tx_wrs_direct++; EQ_LOCK(eq); prev = TAILQ_PREV(cookie, wrq_incomplete_wrs, link); next = TAILQ_NEXT(cookie, link); if (prev == NULL) { MPASS(pidx == eq->dbidx); if (next == NULL || ndesc >= 16) ring_eq_db(wrq->adapter, eq, ndesc); else { MPASS(IDXDIFF(next->pidx, pidx, eq->sidx) == ndesc); next->pidx = pidx; next->ndesc += ndesc; } } else { MPASS(IDXDIFF(pidx, prev->pidx, eq->sidx) == prev->ndesc); prev->ndesc += ndesc; } TAILQ_REMOVE(&wrq->incomplete_wrs, cookie, link); if (TAILQ_EMPTY(&wrq->incomplete_wrs) && !STAILQ_EMPTY(&wrq->wr_list)) drain_wrq_wr_list(sc, wrq); #ifdef INVARIANTS if (TAILQ_EMPTY(&wrq->incomplete_wrs)) { /* Doorbell must have caught up to the pidx. */ MPASS(wrq->eq.pidx == wrq->eq.dbidx); } #endif EQ_UNLOCK(eq); } static u_int can_resume_eth_tx(struct mp_ring *r) { struct sge_eq *eq = r->cookie; return (total_available_tx_desc(eq) > eq->sidx / 8); } static inline int cannot_use_txpkts(struct mbuf *m) { /* maybe put a GL limit too, to avoid silliness? */ return (needs_tso(m)); } /* * r->items[cidx] to r->items[pidx], with a wraparound at r->size, are ready to * be consumed. Return the actual number consumed. 0 indicates a stall. */ static u_int eth_tx(struct mp_ring *r, u_int cidx, u_int pidx) { struct sge_txq *txq = r->cookie; struct sge_eq *eq = &txq->eq; struct ifnet *ifp = txq->ifp; struct port_info *pi = (void *)ifp->if_softc; struct adapter *sc = pi->adapter; u_int total, remaining; /* # of packets */ u_int available, dbdiff; /* # of hardware descriptors */ u_int n, next_cidx; struct mbuf *m0, *tail; struct txpkts txp; struct fw_eth_tx_pkts_wr *wr; /* any fw WR struct will do */ remaining = IDXDIFF(pidx, cidx, r->size); MPASS(remaining > 0); /* Must not be called without work to do. */ total = 0; TXQ_LOCK(txq); if (__predict_false((eq->flags & EQ_ENABLED) == 0)) { while (cidx != pidx) { m0 = r->items[cidx]; m_freem(m0); if (++cidx == r->size) cidx = 0; } reclaim_tx_descs(txq, 2048); total = remaining; goto done; } /* How many hardware descriptors do we have readily available. */ if (eq->pidx == eq->cidx) available = eq->sidx - 1; else available = IDXDIFF(eq->cidx, eq->pidx, eq->sidx) - 1; dbdiff = IDXDIFF(eq->pidx, eq->dbidx, eq->sidx); while (remaining > 0) { m0 = r->items[cidx]; M_ASSERTPKTHDR(m0); MPASS(m0->m_nextpkt == NULL); if (available < SGE_MAX_WR_NDESC) { available += reclaim_tx_descs(txq, 64); if (available < howmany(mbuf_len16(m0), EQ_ESIZE / 16)) break; /* out of descriptors */ } next_cidx = cidx + 1; if (__predict_false(next_cidx == r->size)) next_cidx = 0; wr = (void *)&eq->desc[eq->pidx]; if (remaining > 1 && try_txpkts(m0, r->items[next_cidx], &txp, available) == 0) { /* pkts at cidx, next_cidx should both be in txp. */ MPASS(txp.npkt == 2); tail = r->items[next_cidx]; MPASS(tail->m_nextpkt == NULL); ETHER_BPF_MTAP(ifp, m0); ETHER_BPF_MTAP(ifp, tail); m0->m_nextpkt = tail; if (__predict_false(++next_cidx == r->size)) next_cidx = 0; while (next_cidx != pidx) { if (add_to_txpkts(r->items[next_cidx], &txp, available) != 0) break; tail->m_nextpkt = r->items[next_cidx]; tail = tail->m_nextpkt; ETHER_BPF_MTAP(ifp, tail); if (__predict_false(++next_cidx == r->size)) next_cidx = 0; } n = write_txpkts_wr(txq, wr, m0, &txp, available); total += txp.npkt; remaining -= txp.npkt; } else { total++; remaining--; n = write_txpkt_wr(txq, (void *)wr, m0, available); ETHER_BPF_MTAP(ifp, m0); } MPASS(n >= 1 && n <= available && n <= SGE_MAX_WR_NDESC); available -= n; dbdiff += n; IDXINCR(eq->pidx, n, eq->sidx); if (total_available_tx_desc(eq) < eq->sidx / 4 && atomic_cmpset_int(&eq->equiq, 0, 1)) { wr->equiq_to_len16 |= htobe32(F_FW_WR_EQUIQ | F_FW_WR_EQUEQ); eq->equeqidx = eq->pidx; } else if (IDXDIFF(eq->pidx, eq->equeqidx, eq->sidx) >= 32) { wr->equiq_to_len16 |= htobe32(F_FW_WR_EQUEQ); eq->equeqidx = eq->pidx; } if (dbdiff >= 16 && remaining >= 4) { ring_eq_db(sc, eq, dbdiff); available += reclaim_tx_descs(txq, 4 * dbdiff); dbdiff = 0; } cidx = next_cidx; } if (dbdiff != 0) { ring_eq_db(sc, eq, dbdiff); reclaim_tx_descs(txq, 32); } done: TXQ_UNLOCK(txq); return (total); } static inline void init_iq(struct sge_iq *iq, struct adapter *sc, int tmr_idx, int pktc_idx, int qsize) { KASSERT(tmr_idx >= 0 && tmr_idx < SGE_NTIMERS, ("%s: bad tmr_idx %d", __func__, tmr_idx)); KASSERT(pktc_idx < SGE_NCOUNTERS, /* -ve is ok, means don't use */ ("%s: bad pktc_idx %d", __func__, pktc_idx)); iq->flags = 0; iq->adapter = sc; iq->intr_params = V_QINTR_TIMER_IDX(tmr_idx); iq->intr_pktc_idx = SGE_NCOUNTERS - 1; if (pktc_idx >= 0) { iq->intr_params |= F_QINTR_CNT_EN; iq->intr_pktc_idx = pktc_idx; } iq->qsize = roundup2(qsize, 16); /* See FW_IQ_CMD/iqsize */ iq->sidx = iq->qsize - spg_len / IQ_ESIZE; } static inline void init_fl(struct adapter *sc, struct sge_fl *fl, int qsize, int maxp, char *name) { fl->qsize = qsize; fl->sidx = qsize - spg_len / EQ_ESIZE; strlcpy(fl->lockname, name, sizeof(fl->lockname)); if (sc->flags & BUF_PACKING_OK && ((!is_t4(sc) && buffer_packing) || /* T5+: enabled unless 0 */ (is_t4(sc) && buffer_packing == 1)))/* T4: disabled unless 1 */ fl->flags |= FL_BUF_PACKING; find_best_refill_source(sc, fl, maxp); find_safe_refill_source(sc, fl); } static inline void init_eq(struct sge_eq *eq, int eqtype, int qsize, uint8_t tx_chan, uint16_t iqid, char *name) { KASSERT(tx_chan < NCHAN, ("%s: bad tx channel %d", __func__, tx_chan)); KASSERT(eqtype <= EQ_TYPEMASK, ("%s: bad qtype %d", __func__, eqtype)); eq->flags = eqtype & EQ_TYPEMASK; eq->tx_chan = tx_chan; eq->iqid = iqid; eq->sidx = qsize - spg_len / EQ_ESIZE; strlcpy(eq->lockname, name, sizeof(eq->lockname)); } static int alloc_ring(struct adapter *sc, size_t len, bus_dma_tag_t *tag, bus_dmamap_t *map, bus_addr_t *pa, void **va) { int rc; rc = bus_dma_tag_create(sc->dmat, 512, 0, BUS_SPACE_MAXADDR, BUS_SPACE_MAXADDR, NULL, NULL, len, 1, len, 0, NULL, NULL, tag); if (rc != 0) { device_printf(sc->dev, "cannot allocate DMA tag: %d\n", rc); goto done; } rc = bus_dmamem_alloc(*tag, va, BUS_DMA_WAITOK | BUS_DMA_COHERENT | BUS_DMA_ZERO, map); if (rc != 0) { device_printf(sc->dev, "cannot allocate DMA memory: %d\n", rc); goto done; } rc = bus_dmamap_load(*tag, *map, *va, len, oneseg_dma_callback, pa, 0); if (rc != 0) { device_printf(sc->dev, "cannot load DMA map: %d\n", rc); goto done; } done: if (rc) free_ring(sc, *tag, *map, *pa, *va); return (rc); } static int free_ring(struct adapter *sc, bus_dma_tag_t tag, bus_dmamap_t map, bus_addr_t pa, void *va) { if (pa) bus_dmamap_unload(tag, map); if (va) bus_dmamem_free(tag, va, map); if (tag) bus_dma_tag_destroy(tag); return (0); } /* * Allocates the ring for an ingress queue and an optional freelist. If the * freelist is specified it will be allocated and then associated with the * ingress queue. * * Returns errno on failure. Resources allocated up to that point may still be * allocated. Caller is responsible for cleanup in case this function fails. * * If the ingress queue will take interrupts directly (iq->flags & IQ_INTR) then * the intr_idx specifies the vector, starting from 0. Otherwise it specifies * the abs_id of the ingress queue to which its interrupts should be forwarded. */ static int alloc_iq_fl(struct port_info *pi, struct sge_iq *iq, struct sge_fl *fl, int intr_idx, int cong) { int rc, i, cntxt_id; size_t len; struct fw_iq_cmd c; struct adapter *sc = iq->adapter; __be32 v = 0; len = iq->qsize * IQ_ESIZE; rc = alloc_ring(sc, len, &iq->desc_tag, &iq->desc_map, &iq->ba, (void **)&iq->desc); if (rc != 0) return (rc); bzero(&c, sizeof(c)); c.op_to_vfn = htobe32(V_FW_CMD_OP(FW_IQ_CMD) | F_FW_CMD_REQUEST | F_FW_CMD_WRITE | F_FW_CMD_EXEC | V_FW_IQ_CMD_PFN(sc->pf) | V_FW_IQ_CMD_VFN(0)); c.alloc_to_len16 = htobe32(F_FW_IQ_CMD_ALLOC | F_FW_IQ_CMD_IQSTART | FW_LEN16(c)); /* Special handling for firmware event queue */ if (iq == &sc->sge.fwq) v |= F_FW_IQ_CMD_IQASYNCH; if (iq->flags & IQ_INTR) { KASSERT(intr_idx < sc->intr_count, ("%s: invalid direct intr_idx %d", __func__, intr_idx)); } else v |= F_FW_IQ_CMD_IQANDST; v |= V_FW_IQ_CMD_IQANDSTINDEX(intr_idx); c.type_to_iqandstindex = htobe32(v | V_FW_IQ_CMD_TYPE(FW_IQ_TYPE_FL_INT_CAP) | V_FW_IQ_CMD_VIID(pi->viid) | V_FW_IQ_CMD_IQANUD(X_UPDATEDELIVERY_INTERRUPT)); c.iqdroprss_to_iqesize = htobe16(V_FW_IQ_CMD_IQPCIECH(pi->tx_chan) | F_FW_IQ_CMD_IQGTSMODE | V_FW_IQ_CMD_IQINTCNTTHRESH(iq->intr_pktc_idx) | V_FW_IQ_CMD_IQESIZE(ilog2(IQ_ESIZE) - 4)); c.iqsize = htobe16(iq->qsize); c.iqaddr = htobe64(iq->ba); if (cong >= 0) c.iqns_to_fl0congen = htobe32(F_FW_IQ_CMD_IQFLINTCONGEN); if (fl) { mtx_init(&fl->fl_lock, fl->lockname, NULL, MTX_DEF); len = fl->qsize * EQ_ESIZE; rc = alloc_ring(sc, len, &fl->desc_tag, &fl->desc_map, &fl->ba, (void **)&fl->desc); if (rc) return (rc); /* Allocate space for one software descriptor per buffer. */ rc = alloc_fl_sdesc(fl); if (rc != 0) { device_printf(sc->dev, "failed to setup fl software descriptors: %d\n", rc); return (rc); } if (fl->flags & FL_BUF_PACKING) { fl->lowat = roundup2(sc->sge.fl_starve_threshold2, 8); fl->buf_boundary = sc->sge.pack_boundary; } else { fl->lowat = roundup2(sc->sge.fl_starve_threshold, 8); fl->buf_boundary = 16; } if (fl_pad && fl->buf_boundary < sc->sge.pad_boundary) fl->buf_boundary = sc->sge.pad_boundary; c.iqns_to_fl0congen |= htobe32(V_FW_IQ_CMD_FL0HOSTFCMODE(X_HOSTFCMODE_NONE) | F_FW_IQ_CMD_FL0FETCHRO | F_FW_IQ_CMD_FL0DATARO | (fl_pad ? F_FW_IQ_CMD_FL0PADEN : 0) | (fl->flags & FL_BUF_PACKING ? F_FW_IQ_CMD_FL0PACKEN : 0)); if (cong >= 0) { c.iqns_to_fl0congen |= htobe32(V_FW_IQ_CMD_FL0CNGCHMAP(cong) | F_FW_IQ_CMD_FL0CONGCIF | F_FW_IQ_CMD_FL0CONGEN); } c.fl0dcaen_to_fl0cidxfthresh = - htobe16(V_FW_IQ_CMD_FL0FBMIN(X_FETCHBURSTMIN_64B) | + htobe16(V_FW_IQ_CMD_FL0FBMIN(X_FETCHBURSTMIN_128B) | V_FW_IQ_CMD_FL0FBMAX(X_FETCHBURSTMAX_512B)); c.fl0size = htobe16(fl->qsize); c.fl0addr = htobe64(fl->ba); } rc = -t4_wr_mbox(sc, sc->mbox, &c, sizeof(c), &c); if (rc != 0) { device_printf(sc->dev, "failed to create ingress queue: %d\n", rc); return (rc); } iq->cidx = 0; iq->gen = F_RSPD_GEN; iq->intr_next = iq->intr_params; iq->cntxt_id = be16toh(c.iqid); iq->abs_id = be16toh(c.physiqid); iq->flags |= IQ_ALLOCATED; cntxt_id = iq->cntxt_id - sc->sge.iq_start; if (cntxt_id >= sc->sge.niq) { panic ("%s: iq->cntxt_id (%d) more than the max (%d)", __func__, cntxt_id, sc->sge.niq - 1); } sc->sge.iqmap[cntxt_id] = iq; if (fl) { u_int qid; iq->flags |= IQ_HAS_FL; fl->cntxt_id = be16toh(c.fl0id); fl->pidx = fl->cidx = 0; cntxt_id = fl->cntxt_id - sc->sge.eq_start; if (cntxt_id >= sc->sge.neq) { panic("%s: fl->cntxt_id (%d) more than the max (%d)", __func__, cntxt_id, sc->sge.neq - 1); } sc->sge.eqmap[cntxt_id] = (void *)fl; qid = fl->cntxt_id; if (isset(&sc->doorbells, DOORBELL_UDB)) { uint32_t s_qpp = sc->sge.eq_s_qpp; uint32_t mask = (1 << s_qpp) - 1; volatile uint8_t *udb; udb = sc->udbs_base + UDBS_DB_OFFSET; udb += (qid >> s_qpp) << PAGE_SHIFT; qid &= mask; if (qid < PAGE_SIZE / UDBS_SEG_SIZE) { udb += qid << UDBS_SEG_SHIFT; qid = 0; } fl->udb = (volatile void *)udb; } fl->dbval = F_DBPRIO | V_QID(qid); if (is_t5(sc)) fl->dbval |= F_DBTYPE; FL_LOCK(fl); /* Enough to make sure the SGE doesn't think it's starved */ refill_fl(sc, fl, fl->lowat); FL_UNLOCK(fl); } if (is_t5(sc) && cong >= 0) { uint32_t param, val; param = V_FW_PARAMS_MNEM(FW_PARAMS_MNEM_DMAQ) | V_FW_PARAMS_PARAM_X(FW_PARAMS_PARAM_DMAQ_CONM_CTXT) | V_FW_PARAMS_PARAM_YZ(iq->cntxt_id); if (cong == 0) val = 1 << 19; else { val = 2 << 19; for (i = 0; i < 4; i++) { if (cong & (1 << i)) val |= 1 << (i << 2); } } rc = -t4_set_params(sc, sc->mbox, sc->pf, 0, 1, ¶m, &val); if (rc != 0) { /* report error but carry on */ device_printf(sc->dev, "failed to set congestion manager context for " "ingress queue %d: %d\n", iq->cntxt_id, rc); } } /* Enable IQ interrupts */ atomic_store_rel_int(&iq->state, IQS_IDLE); t4_write_reg(sc, MYPF_REG(A_SGE_PF_GTS), V_SEINTARM(iq->intr_params) | V_INGRESSQID(iq->cntxt_id)); return (0); } static int free_iq_fl(struct port_info *pi, struct sge_iq *iq, struct sge_fl *fl) { int rc; struct adapter *sc = iq->adapter; device_t dev; if (sc == NULL) return (0); /* nothing to do */ dev = pi ? pi->dev : sc->dev; if (iq->flags & IQ_ALLOCATED) { rc = -t4_iq_free(sc, sc->mbox, sc->pf, 0, FW_IQ_TYPE_FL_INT_CAP, iq->cntxt_id, fl ? fl->cntxt_id : 0xffff, 0xffff); if (rc != 0) { device_printf(dev, "failed to free queue %p: %d\n", iq, rc); return (rc); } iq->flags &= ~IQ_ALLOCATED; } free_ring(sc, iq->desc_tag, iq->desc_map, iq->ba, iq->desc); bzero(iq, sizeof(*iq)); if (fl) { free_ring(sc, fl->desc_tag, fl->desc_map, fl->ba, fl->desc); if (fl->sdesc) free_fl_sdesc(sc, fl); if (mtx_initialized(&fl->fl_lock)) mtx_destroy(&fl->fl_lock); bzero(fl, sizeof(*fl)); } return (0); } static void add_fl_sysctls(struct sysctl_ctx_list *ctx, struct sysctl_oid *oid, struct sge_fl *fl) { struct sysctl_oid_list *children = SYSCTL_CHILDREN(oid); oid = SYSCTL_ADD_NODE(ctx, children, OID_AUTO, "fl", CTLFLAG_RD, NULL, "freelist"); children = SYSCTL_CHILDREN(oid); SYSCTL_ADD_PROC(ctx, children, OID_AUTO, "cntxt_id", CTLTYPE_INT | CTLFLAG_RD, &fl->cntxt_id, 0, sysctl_uint16, "I", "SGE context id of the freelist"); SYSCTL_ADD_UINT(ctx, children, OID_AUTO, "padding", CTLFLAG_RD, NULL, fl_pad ? 1 : 0, "padding enabled"); SYSCTL_ADD_UINT(ctx, children, OID_AUTO, "packing", CTLFLAG_RD, NULL, fl->flags & FL_BUF_PACKING ? 1 : 0, "packing enabled"); SYSCTL_ADD_UINT(ctx, children, OID_AUTO, "cidx", CTLFLAG_RD, &fl->cidx, 0, "consumer index"); if (fl->flags & FL_BUF_PACKING) { SYSCTL_ADD_UINT(ctx, children, OID_AUTO, "rx_offset", CTLFLAG_RD, &fl->rx_offset, 0, "packing rx offset"); } SYSCTL_ADD_UINT(ctx, children, OID_AUTO, "pidx", CTLFLAG_RD, &fl->pidx, 0, "producer index"); SYSCTL_ADD_UQUAD(ctx, children, OID_AUTO, "mbuf_allocated", CTLFLAG_RD, &fl->mbuf_allocated, "# of mbuf allocated"); SYSCTL_ADD_UQUAD(ctx, children, OID_AUTO, "mbuf_inlined", CTLFLAG_RD, &fl->mbuf_inlined, "# of mbuf inlined in clusters"); SYSCTL_ADD_UQUAD(ctx, children, OID_AUTO, "cluster_allocated", CTLFLAG_RD, &fl->cl_allocated, "# of clusters allocated"); SYSCTL_ADD_UQUAD(ctx, children, OID_AUTO, "cluster_recycled", CTLFLAG_RD, &fl->cl_recycled, "# of clusters recycled"); SYSCTL_ADD_UQUAD(ctx, children, OID_AUTO, "cluster_fast_recycled", CTLFLAG_RD, &fl->cl_fast_recycled, "# of clusters recycled (fast)"); } static int alloc_fwq(struct adapter *sc) { int rc, intr_idx; struct sge_iq *fwq = &sc->sge.fwq; struct sysctl_oid *oid = device_get_sysctl_tree(sc->dev); struct sysctl_oid_list *children = SYSCTL_CHILDREN(oid); init_iq(fwq, sc, 0, 0, FW_IQ_QSIZE); fwq->flags |= IQ_INTR; /* always */ intr_idx = sc->intr_count > 1 ? 1 : 0; rc = alloc_iq_fl(sc->port[0], fwq, NULL, intr_idx, -1); if (rc != 0) { device_printf(sc->dev, "failed to create firmware event queue: %d\n", rc); return (rc); } oid = SYSCTL_ADD_NODE(&sc->ctx, children, OID_AUTO, "fwq", CTLFLAG_RD, NULL, "firmware event queue"); children = SYSCTL_CHILDREN(oid); SYSCTL_ADD_PROC(&sc->ctx, children, OID_AUTO, "abs_id", CTLTYPE_INT | CTLFLAG_RD, &fwq->abs_id, 0, sysctl_uint16, "I", "absolute id of the queue"); SYSCTL_ADD_PROC(&sc->ctx, children, OID_AUTO, "cntxt_id", CTLTYPE_INT | CTLFLAG_RD, &fwq->cntxt_id, 0, sysctl_uint16, "I", "SGE context id of the queue"); SYSCTL_ADD_PROC(&sc->ctx, children, OID_AUTO, "cidx", CTLTYPE_INT | CTLFLAG_RD, &fwq->cidx, 0, sysctl_uint16, "I", "consumer index"); return (0); } static int free_fwq(struct adapter *sc) { return free_iq_fl(NULL, &sc->sge.fwq, NULL); } static int alloc_mgmtq(struct adapter *sc) { int rc; struct sge_wrq *mgmtq = &sc->sge.mgmtq; char name[16]; struct sysctl_oid *oid = device_get_sysctl_tree(sc->dev); struct sysctl_oid_list *children = SYSCTL_CHILDREN(oid); oid = SYSCTL_ADD_NODE(&sc->ctx, children, OID_AUTO, "mgmtq", CTLFLAG_RD, NULL, "management queue"); snprintf(name, sizeof(name), "%s mgmtq", device_get_nameunit(sc->dev)); init_eq(&mgmtq->eq, EQ_CTRL, CTRL_EQ_QSIZE, sc->port[0]->tx_chan, sc->sge.fwq.cntxt_id, name); rc = alloc_wrq(sc, NULL, mgmtq, oid); if (rc != 0) { device_printf(sc->dev, "failed to create management queue: %d\n", rc); return (rc); } return (0); } static int free_mgmtq(struct adapter *sc) { return free_wrq(sc, &sc->sge.mgmtq); } int tnl_cong(struct port_info *pi) { if (cong_drop == -1) return (-1); else if (cong_drop == 1) return (0); else return (pi->rx_chan_map); } static int alloc_rxq(struct port_info *pi, struct sge_rxq *rxq, int intr_idx, int idx, struct sysctl_oid *oid) { int rc; struct sysctl_oid_list *children; char name[16]; rc = alloc_iq_fl(pi, &rxq->iq, &rxq->fl, intr_idx, tnl_cong(pi)); if (rc != 0) return (rc); /* * The freelist is just barely above the starvation threshold right now, * fill it up a bit more. */ FL_LOCK(&rxq->fl); refill_fl(pi->adapter, &rxq->fl, 128); FL_UNLOCK(&rxq->fl); #if defined(INET) || defined(INET6) rc = tcp_lro_init(&rxq->lro); if (rc != 0) return (rc); rxq->lro.ifp = pi->ifp; /* also indicates LRO init'ed */ if (pi->ifp->if_capenable & IFCAP_LRO) rxq->iq.flags |= IQ_LRO_ENABLED; #endif rxq->ifp = pi->ifp; children = SYSCTL_CHILDREN(oid); snprintf(name, sizeof(name), "%d", idx); oid = SYSCTL_ADD_NODE(&pi->ctx, children, OID_AUTO, name, CTLFLAG_RD, NULL, "rx queue"); children = SYSCTL_CHILDREN(oid); SYSCTL_ADD_PROC(&pi->ctx, children, OID_AUTO, "abs_id", CTLTYPE_INT | CTLFLAG_RD, &rxq->iq.abs_id, 0, sysctl_uint16, "I", "absolute id of the queue"); SYSCTL_ADD_PROC(&pi->ctx, children, OID_AUTO, "cntxt_id", CTLTYPE_INT | CTLFLAG_RD, &rxq->iq.cntxt_id, 0, sysctl_uint16, "I", "SGE context id of the queue"); SYSCTL_ADD_PROC(&pi->ctx, children, OID_AUTO, "cidx", CTLTYPE_INT | CTLFLAG_RD, &rxq->iq.cidx, 0, sysctl_uint16, "I", "consumer index"); #if defined(INET) || defined(INET6) SYSCTL_ADD_INT(&pi->ctx, children, OID_AUTO, "lro_queued", CTLFLAG_RD, &rxq->lro.lro_queued, 0, NULL); SYSCTL_ADD_INT(&pi->ctx, children, OID_AUTO, "lro_flushed", CTLFLAG_RD, &rxq->lro.lro_flushed, 0, NULL); #endif SYSCTL_ADD_UQUAD(&pi->ctx, children, OID_AUTO, "rxcsum", CTLFLAG_RD, &rxq->rxcsum, "# of times hardware assisted with checksum"); SYSCTL_ADD_UQUAD(&pi->ctx, children, OID_AUTO, "vlan_extraction", CTLFLAG_RD, &rxq->vlan_extraction, "# of times hardware extracted 802.1Q tag"); add_fl_sysctls(&pi->ctx, oid, &rxq->fl); return (rc); } static int free_rxq(struct port_info *pi, struct sge_rxq *rxq) { int rc; #if defined(INET) || defined(INET6) if (rxq->lro.ifp) { tcp_lro_free(&rxq->lro); rxq->lro.ifp = NULL; } #endif rc = free_iq_fl(pi, &rxq->iq, &rxq->fl); if (rc == 0) bzero(rxq, sizeof(*rxq)); return (rc); } #ifdef TCP_OFFLOAD static int alloc_ofld_rxq(struct port_info *pi, struct sge_ofld_rxq *ofld_rxq, int intr_idx, int idx, struct sysctl_oid *oid) { int rc; struct sysctl_oid_list *children; char name[16]; rc = alloc_iq_fl(pi, &ofld_rxq->iq, &ofld_rxq->fl, intr_idx, pi->rx_chan_map); if (rc != 0) return (rc); children = SYSCTL_CHILDREN(oid); snprintf(name, sizeof(name), "%d", idx); oid = SYSCTL_ADD_NODE(&pi->ctx, children, OID_AUTO, name, CTLFLAG_RD, NULL, "rx queue"); children = SYSCTL_CHILDREN(oid); SYSCTL_ADD_PROC(&pi->ctx, children, OID_AUTO, "abs_id", CTLTYPE_INT | CTLFLAG_RD, &ofld_rxq->iq.abs_id, 0, sysctl_uint16, "I", "absolute id of the queue"); SYSCTL_ADD_PROC(&pi->ctx, children, OID_AUTO, "cntxt_id", CTLTYPE_INT | CTLFLAG_RD, &ofld_rxq->iq.cntxt_id, 0, sysctl_uint16, "I", "SGE context id of the queue"); SYSCTL_ADD_PROC(&pi->ctx, children, OID_AUTO, "cidx", CTLTYPE_INT | CTLFLAG_RD, &ofld_rxq->iq.cidx, 0, sysctl_uint16, "I", "consumer index"); add_fl_sysctls(&pi->ctx, oid, &ofld_rxq->fl); return (rc); } static int free_ofld_rxq(struct port_info *pi, struct sge_ofld_rxq *ofld_rxq) { int rc; rc = free_iq_fl(pi, &ofld_rxq->iq, &ofld_rxq->fl); if (rc == 0) bzero(ofld_rxq, sizeof(*ofld_rxq)); return (rc); } #endif #ifdef DEV_NETMAP static int alloc_nm_rxq(struct port_info *pi, struct sge_nm_rxq *nm_rxq, int intr_idx, int idx, struct sysctl_oid *oid) { int rc; struct sysctl_oid_list *children; struct sysctl_ctx_list *ctx; char name[16]; size_t len; struct adapter *sc = pi->adapter; struct netmap_adapter *na = NA(pi->nm_ifp); MPASS(na != NULL); len = pi->qsize_rxq * IQ_ESIZE; rc = alloc_ring(sc, len, &nm_rxq->iq_desc_tag, &nm_rxq->iq_desc_map, &nm_rxq->iq_ba, (void **)&nm_rxq->iq_desc); if (rc != 0) return (rc); len = na->num_rx_desc * EQ_ESIZE + spg_len; rc = alloc_ring(sc, len, &nm_rxq->fl_desc_tag, &nm_rxq->fl_desc_map, &nm_rxq->fl_ba, (void **)&nm_rxq->fl_desc); if (rc != 0) return (rc); nm_rxq->pi = pi; nm_rxq->nid = idx; nm_rxq->iq_cidx = 0; nm_rxq->iq_sidx = pi->qsize_rxq - spg_len / IQ_ESIZE; nm_rxq->iq_gen = F_RSPD_GEN; nm_rxq->fl_pidx = nm_rxq->fl_cidx = 0; nm_rxq->fl_sidx = na->num_rx_desc; nm_rxq->intr_idx = intr_idx; ctx = &pi->ctx; children = SYSCTL_CHILDREN(oid); snprintf(name, sizeof(name), "%d", idx); oid = SYSCTL_ADD_NODE(ctx, children, OID_AUTO, name, CTLFLAG_RD, NULL, "rx queue"); children = SYSCTL_CHILDREN(oid); SYSCTL_ADD_PROC(ctx, children, OID_AUTO, "abs_id", CTLTYPE_INT | CTLFLAG_RD, &nm_rxq->iq_abs_id, 0, sysctl_uint16, "I", "absolute id of the queue"); SYSCTL_ADD_PROC(ctx, children, OID_AUTO, "cntxt_id", CTLTYPE_INT | CTLFLAG_RD, &nm_rxq->iq_cntxt_id, 0, sysctl_uint16, "I", "SGE context id of the queue"); SYSCTL_ADD_PROC(ctx, children, OID_AUTO, "cidx", CTLTYPE_INT | CTLFLAG_RD, &nm_rxq->iq_cidx, 0, sysctl_uint16, "I", "consumer index"); children = SYSCTL_CHILDREN(oid); oid = SYSCTL_ADD_NODE(ctx, children, OID_AUTO, "fl", CTLFLAG_RD, NULL, "freelist"); children = SYSCTL_CHILDREN(oid); SYSCTL_ADD_PROC(ctx, children, OID_AUTO, "cntxt_id", CTLTYPE_INT | CTLFLAG_RD, &nm_rxq->fl_cntxt_id, 0, sysctl_uint16, "I", "SGE context id of the freelist"); SYSCTL_ADD_UINT(ctx, children, OID_AUTO, "cidx", CTLFLAG_RD, &nm_rxq->fl_cidx, 0, "consumer index"); SYSCTL_ADD_UINT(ctx, children, OID_AUTO, "pidx", CTLFLAG_RD, &nm_rxq->fl_pidx, 0, "producer index"); return (rc); } static int free_nm_rxq(struct port_info *pi, struct sge_nm_rxq *nm_rxq) { struct adapter *sc = pi->adapter; free_ring(sc, nm_rxq->iq_desc_tag, nm_rxq->iq_desc_map, nm_rxq->iq_ba, nm_rxq->iq_desc); free_ring(sc, nm_rxq->fl_desc_tag, nm_rxq->fl_desc_map, nm_rxq->fl_ba, nm_rxq->fl_desc); return (0); } static int alloc_nm_txq(struct port_info *pi, struct sge_nm_txq *nm_txq, int iqidx, int idx, struct sysctl_oid *oid) { int rc; size_t len; struct adapter *sc = pi->adapter; struct netmap_adapter *na = NA(pi->nm_ifp); char name[16]; struct sysctl_oid_list *children = SYSCTL_CHILDREN(oid); len = na->num_tx_desc * EQ_ESIZE + spg_len; rc = alloc_ring(sc, len, &nm_txq->desc_tag, &nm_txq->desc_map, &nm_txq->ba, (void **)&nm_txq->desc); if (rc) return (rc); nm_txq->pidx = nm_txq->cidx = 0; nm_txq->sidx = na->num_tx_desc; nm_txq->nid = idx; nm_txq->iqidx = iqidx; nm_txq->cpl_ctrl0 = htobe32(V_TXPKT_OPCODE(CPL_TX_PKT) | V_TXPKT_INTF(pi->tx_chan) | V_TXPKT_PF(sc->pf)); snprintf(name, sizeof(name), "%d", idx); oid = SYSCTL_ADD_NODE(&pi->ctx, children, OID_AUTO, name, CTLFLAG_RD, NULL, "netmap tx queue"); children = SYSCTL_CHILDREN(oid); SYSCTL_ADD_UINT(&pi->ctx, children, OID_AUTO, "cntxt_id", CTLFLAG_RD, &nm_txq->cntxt_id, 0, "SGE context id of the queue"); SYSCTL_ADD_PROC(&pi->ctx, children, OID_AUTO, "cidx", CTLTYPE_INT | CTLFLAG_RD, &nm_txq->cidx, 0, sysctl_uint16, "I", "consumer index"); SYSCTL_ADD_PROC(&pi->ctx, children, OID_AUTO, "pidx", CTLTYPE_INT | CTLFLAG_RD, &nm_txq->pidx, 0, sysctl_uint16, "I", "producer index"); return (rc); } static int free_nm_txq(struct port_info *pi, struct sge_nm_txq *nm_txq) { struct adapter *sc = pi->adapter; free_ring(sc, nm_txq->desc_tag, nm_txq->desc_map, nm_txq->ba, nm_txq->desc); return (0); } #endif static int ctrl_eq_alloc(struct adapter *sc, struct sge_eq *eq) { int rc, cntxt_id; struct fw_eq_ctrl_cmd c; int qsize = eq->sidx + spg_len / EQ_ESIZE; bzero(&c, sizeof(c)); c.op_to_vfn = htobe32(V_FW_CMD_OP(FW_EQ_CTRL_CMD) | F_FW_CMD_REQUEST | F_FW_CMD_WRITE | F_FW_CMD_EXEC | V_FW_EQ_CTRL_CMD_PFN(sc->pf) | V_FW_EQ_CTRL_CMD_VFN(0)); c.alloc_to_len16 = htobe32(F_FW_EQ_CTRL_CMD_ALLOC | F_FW_EQ_CTRL_CMD_EQSTART | FW_LEN16(c)); c.cmpliqid_eqid = htonl(V_FW_EQ_CTRL_CMD_CMPLIQID(eq->iqid)); c.physeqid_pkd = htobe32(0); c.fetchszm_to_iqid = htobe32(V_FW_EQ_CTRL_CMD_HOSTFCMODE(X_HOSTFCMODE_NONE) | V_FW_EQ_CTRL_CMD_PCIECHN(eq->tx_chan) | F_FW_EQ_CTRL_CMD_FETCHRO | V_FW_EQ_CTRL_CMD_IQID(eq->iqid)); c.dcaen_to_eqsize = htobe32(V_FW_EQ_CTRL_CMD_FBMIN(X_FETCHBURSTMIN_64B) | V_FW_EQ_CTRL_CMD_FBMAX(X_FETCHBURSTMAX_512B) | V_FW_EQ_CTRL_CMD_EQSIZE(qsize)); c.eqaddr = htobe64(eq->ba); rc = -t4_wr_mbox(sc, sc->mbox, &c, sizeof(c), &c); if (rc != 0) { device_printf(sc->dev, "failed to create control queue %d: %d\n", eq->tx_chan, rc); return (rc); } eq->flags |= EQ_ALLOCATED; eq->cntxt_id = G_FW_EQ_CTRL_CMD_EQID(be32toh(c.cmpliqid_eqid)); cntxt_id = eq->cntxt_id - sc->sge.eq_start; if (cntxt_id >= sc->sge.neq) panic("%s: eq->cntxt_id (%d) more than the max (%d)", __func__, cntxt_id, sc->sge.neq - 1); sc->sge.eqmap[cntxt_id] = eq; return (rc); } static int eth_eq_alloc(struct adapter *sc, struct port_info *pi, struct sge_eq *eq) { int rc, cntxt_id; struct fw_eq_eth_cmd c; int qsize = eq->sidx + spg_len / EQ_ESIZE; bzero(&c, sizeof(c)); c.op_to_vfn = htobe32(V_FW_CMD_OP(FW_EQ_ETH_CMD) | F_FW_CMD_REQUEST | F_FW_CMD_WRITE | F_FW_CMD_EXEC | V_FW_EQ_ETH_CMD_PFN(sc->pf) | V_FW_EQ_ETH_CMD_VFN(0)); c.alloc_to_len16 = htobe32(F_FW_EQ_ETH_CMD_ALLOC | F_FW_EQ_ETH_CMD_EQSTART | FW_LEN16(c)); c.autoequiqe_to_viid = htobe32(F_FW_EQ_ETH_CMD_AUTOEQUIQE | F_FW_EQ_ETH_CMD_AUTOEQUEQE | V_FW_EQ_ETH_CMD_VIID(pi->viid)); c.fetchszm_to_iqid = htobe32(V_FW_EQ_ETH_CMD_HOSTFCMODE(X_HOSTFCMODE_NONE) | V_FW_EQ_ETH_CMD_PCIECHN(eq->tx_chan) | F_FW_EQ_ETH_CMD_FETCHRO | V_FW_EQ_ETH_CMD_IQID(eq->iqid)); c.dcaen_to_eqsize = htobe32(V_FW_EQ_ETH_CMD_FBMIN(X_FETCHBURSTMIN_64B) | V_FW_EQ_ETH_CMD_FBMAX(X_FETCHBURSTMAX_512B) | V_FW_EQ_ETH_CMD_EQSIZE(qsize)); c.eqaddr = htobe64(eq->ba); rc = -t4_wr_mbox(sc, sc->mbox, &c, sizeof(c), &c); if (rc != 0) { device_printf(pi->dev, "failed to create Ethernet egress queue: %d\n", rc); return (rc); } eq->flags |= EQ_ALLOCATED; eq->cntxt_id = G_FW_EQ_ETH_CMD_EQID(be32toh(c.eqid_pkd)); cntxt_id = eq->cntxt_id - sc->sge.eq_start; if (cntxt_id >= sc->sge.neq) panic("%s: eq->cntxt_id (%d) more than the max (%d)", __func__, cntxt_id, sc->sge.neq - 1); sc->sge.eqmap[cntxt_id] = eq; return (rc); } #ifdef TCP_OFFLOAD static int ofld_eq_alloc(struct adapter *sc, struct port_info *pi, struct sge_eq *eq) { int rc, cntxt_id; struct fw_eq_ofld_cmd c; int qsize = eq->sidx + spg_len / EQ_ESIZE; bzero(&c, sizeof(c)); c.op_to_vfn = htonl(V_FW_CMD_OP(FW_EQ_OFLD_CMD) | F_FW_CMD_REQUEST | F_FW_CMD_WRITE | F_FW_CMD_EXEC | V_FW_EQ_OFLD_CMD_PFN(sc->pf) | V_FW_EQ_OFLD_CMD_VFN(0)); c.alloc_to_len16 = htonl(F_FW_EQ_OFLD_CMD_ALLOC | F_FW_EQ_OFLD_CMD_EQSTART | FW_LEN16(c)); c.fetchszm_to_iqid = htonl(V_FW_EQ_OFLD_CMD_HOSTFCMODE(X_HOSTFCMODE_NONE) | V_FW_EQ_OFLD_CMD_PCIECHN(eq->tx_chan) | F_FW_EQ_OFLD_CMD_FETCHRO | V_FW_EQ_OFLD_CMD_IQID(eq->iqid)); c.dcaen_to_eqsize = htobe32(V_FW_EQ_OFLD_CMD_FBMIN(X_FETCHBURSTMIN_64B) | V_FW_EQ_OFLD_CMD_FBMAX(X_FETCHBURSTMAX_512B) | V_FW_EQ_OFLD_CMD_EQSIZE(qsize)); c.eqaddr = htobe64(eq->ba); rc = -t4_wr_mbox(sc, sc->mbox, &c, sizeof(c), &c); if (rc != 0) { device_printf(pi->dev, "failed to create egress queue for TCP offload: %d\n", rc); return (rc); } eq->flags |= EQ_ALLOCATED; eq->cntxt_id = G_FW_EQ_OFLD_CMD_EQID(be32toh(c.eqid_pkd)); cntxt_id = eq->cntxt_id - sc->sge.eq_start; if (cntxt_id >= sc->sge.neq) panic("%s: eq->cntxt_id (%d) more than the max (%d)", __func__, cntxt_id, sc->sge.neq - 1); sc->sge.eqmap[cntxt_id] = eq; return (rc); } #endif static int alloc_eq(struct adapter *sc, struct port_info *pi, struct sge_eq *eq) { int rc, qsize; size_t len; mtx_init(&eq->eq_lock, eq->lockname, NULL, MTX_DEF); qsize = eq->sidx + spg_len / EQ_ESIZE; len = qsize * EQ_ESIZE; rc = alloc_ring(sc, len, &eq->desc_tag, &eq->desc_map, &eq->ba, (void **)&eq->desc); if (rc) return (rc); eq->pidx = eq->cidx = 0; eq->equeqidx = eq->dbidx = 0; eq->doorbells = sc->doorbells; switch (eq->flags & EQ_TYPEMASK) { case EQ_CTRL: rc = ctrl_eq_alloc(sc, eq); break; case EQ_ETH: rc = eth_eq_alloc(sc, pi, eq); break; #ifdef TCP_OFFLOAD case EQ_OFLD: rc = ofld_eq_alloc(sc, pi, eq); break; #endif default: panic("%s: invalid eq type %d.", __func__, eq->flags & EQ_TYPEMASK); } if (rc != 0) { device_printf(sc->dev, "failed to allocate egress queue(%d): %d\n", eq->flags & EQ_TYPEMASK, rc); } if (isset(&eq->doorbells, DOORBELL_UDB) || isset(&eq->doorbells, DOORBELL_UDBWC) || isset(&eq->doorbells, DOORBELL_WCWR)) { uint32_t s_qpp = sc->sge.eq_s_qpp; uint32_t mask = (1 << s_qpp) - 1; volatile uint8_t *udb; udb = sc->udbs_base + UDBS_DB_OFFSET; udb += (eq->cntxt_id >> s_qpp) << PAGE_SHIFT; /* pg offset */ eq->udb_qid = eq->cntxt_id & mask; /* id in page */ if (eq->udb_qid >= PAGE_SIZE / UDBS_SEG_SIZE) clrbit(&eq->doorbells, DOORBELL_WCWR); else { udb += eq->udb_qid << UDBS_SEG_SHIFT; /* seg offset */ eq->udb_qid = 0; } eq->udb = (volatile void *)udb; } return (rc); } static int free_eq(struct adapter *sc, struct sge_eq *eq) { int rc; if (eq->flags & EQ_ALLOCATED) { switch (eq->flags & EQ_TYPEMASK) { case EQ_CTRL: rc = -t4_ctrl_eq_free(sc, sc->mbox, sc->pf, 0, eq->cntxt_id); break; case EQ_ETH: rc = -t4_eth_eq_free(sc, sc->mbox, sc->pf, 0, eq->cntxt_id); break; #ifdef TCP_OFFLOAD case EQ_OFLD: rc = -t4_ofld_eq_free(sc, sc->mbox, sc->pf, 0, eq->cntxt_id); break; #endif default: panic("%s: invalid eq type %d.", __func__, eq->flags & EQ_TYPEMASK); } if (rc != 0) { device_printf(sc->dev, "failed to free egress queue (%d): %d\n", eq->flags & EQ_TYPEMASK, rc); return (rc); } eq->flags &= ~EQ_ALLOCATED; } free_ring(sc, eq->desc_tag, eq->desc_map, eq->ba, eq->desc); if (mtx_initialized(&eq->eq_lock)) mtx_destroy(&eq->eq_lock); bzero(eq, sizeof(*eq)); return (0); } static int alloc_wrq(struct adapter *sc, struct port_info *pi, struct sge_wrq *wrq, struct sysctl_oid *oid) { int rc; struct sysctl_ctx_list *ctx = pi ? &pi->ctx : &sc->ctx; struct sysctl_oid_list *children = SYSCTL_CHILDREN(oid); rc = alloc_eq(sc, pi, &wrq->eq); if (rc) return (rc); wrq->adapter = sc; TASK_INIT(&wrq->wrq_tx_task, 0, wrq_tx_drain, wrq); TAILQ_INIT(&wrq->incomplete_wrs); STAILQ_INIT(&wrq->wr_list); wrq->nwr_pending = 0; wrq->ndesc_needed = 0; SYSCTL_ADD_UINT(ctx, children, OID_AUTO, "cntxt_id", CTLFLAG_RD, &wrq->eq.cntxt_id, 0, "SGE context id of the queue"); SYSCTL_ADD_PROC(ctx, children, OID_AUTO, "cidx", CTLTYPE_INT | CTLFLAG_RD, &wrq->eq.cidx, 0, sysctl_uint16, "I", "consumer index"); SYSCTL_ADD_PROC(ctx, children, OID_AUTO, "pidx", CTLTYPE_INT | CTLFLAG_RD, &wrq->eq.pidx, 0, sysctl_uint16, "I", "producer index"); SYSCTL_ADD_UQUAD(ctx, children, OID_AUTO, "tx_wrs_direct", CTLFLAG_RD, &wrq->tx_wrs_direct, "# of work requests (direct)"); SYSCTL_ADD_UQUAD(ctx, children, OID_AUTO, "tx_wrs_copied", CTLFLAG_RD, &wrq->tx_wrs_copied, "# of work requests (copied)"); return (rc); } static int free_wrq(struct adapter *sc, struct sge_wrq *wrq) { int rc; rc = free_eq(sc, &wrq->eq); if (rc) return (rc); bzero(wrq, sizeof(*wrq)); return (0); } static int alloc_txq(struct port_info *pi, struct sge_txq *txq, int idx, struct sysctl_oid *oid) { int rc; struct adapter *sc = pi->adapter; struct sge_eq *eq = &txq->eq; char name[16]; struct sysctl_oid_list *children = SYSCTL_CHILDREN(oid); rc = mp_ring_alloc(&txq->r, eq->sidx, txq, eth_tx, can_resume_eth_tx, M_CXGBE, M_WAITOK); if (rc != 0) { device_printf(sc->dev, "failed to allocate mp_ring: %d\n", rc); return (rc); } rc = alloc_eq(sc, pi, eq); if (rc != 0) { mp_ring_free(txq->r); txq->r = NULL; return (rc); } /* Can't fail after this point. */ TASK_INIT(&txq->tx_reclaim_task, 0, tx_reclaim, eq); txq->ifp = pi->ifp; txq->gl = sglist_alloc(TX_SGL_SEGS, M_WAITOK); txq->cpl_ctrl0 = htobe32(V_TXPKT_OPCODE(CPL_TX_PKT) | V_TXPKT_INTF(pi->tx_chan) | V_TXPKT_PF(sc->pf)); txq->sdesc = malloc(eq->sidx * sizeof(struct tx_sdesc), M_CXGBE, M_ZERO | M_WAITOK); snprintf(name, sizeof(name), "%d", idx); oid = SYSCTL_ADD_NODE(&pi->ctx, children, OID_AUTO, name, CTLFLAG_RD, NULL, "tx queue"); children = SYSCTL_CHILDREN(oid); SYSCTL_ADD_UINT(&pi->ctx, children, OID_AUTO, "cntxt_id", CTLFLAG_RD, &eq->cntxt_id, 0, "SGE context id of the queue"); SYSCTL_ADD_PROC(&pi->ctx, children, OID_AUTO, "cidx", CTLTYPE_INT | CTLFLAG_RD, &eq->cidx, 0, sysctl_uint16, "I", "consumer index"); SYSCTL_ADD_PROC(&pi->ctx, children, OID_AUTO, "pidx", CTLTYPE_INT | CTLFLAG_RD, &eq->pidx, 0, sysctl_uint16, "I", "producer index"); SYSCTL_ADD_UQUAD(&pi->ctx, children, OID_AUTO, "txcsum", CTLFLAG_RD, &txq->txcsum, "# of times hardware assisted with checksum"); SYSCTL_ADD_UQUAD(&pi->ctx, children, OID_AUTO, "vlan_insertion", CTLFLAG_RD, &txq->vlan_insertion, "# of times hardware inserted 802.1Q tag"); SYSCTL_ADD_UQUAD(&pi->ctx, children, OID_AUTO, "tso_wrs", CTLFLAG_RD, &txq->tso_wrs, "# of TSO work requests"); SYSCTL_ADD_UQUAD(&pi->ctx, children, OID_AUTO, "imm_wrs", CTLFLAG_RD, &txq->imm_wrs, "# of work requests with immediate data"); SYSCTL_ADD_UQUAD(&pi->ctx, children, OID_AUTO, "sgl_wrs", CTLFLAG_RD, &txq->sgl_wrs, "# of work requests with direct SGL"); SYSCTL_ADD_UQUAD(&pi->ctx, children, OID_AUTO, "txpkt_wrs", CTLFLAG_RD, &txq->txpkt_wrs, "# of txpkt work requests (one pkt/WR)"); SYSCTL_ADD_UQUAD(&pi->ctx, children, OID_AUTO, "txpkts0_wrs", CTLFLAG_RD, &txq->txpkts0_wrs, "# of txpkts (type 0) work requests"); SYSCTL_ADD_UQUAD(&pi->ctx, children, OID_AUTO, "txpkts1_wrs", CTLFLAG_RD, &txq->txpkts1_wrs, "# of txpkts (type 1) work requests"); SYSCTL_ADD_UQUAD(&pi->ctx, children, OID_AUTO, "txpkts0_pkts", CTLFLAG_RD, &txq->txpkts0_pkts, "# of frames tx'd using type0 txpkts work requests"); SYSCTL_ADD_UQUAD(&pi->ctx, children, OID_AUTO, "txpkts1_pkts", CTLFLAG_RD, &txq->txpkts1_pkts, "# of frames tx'd using type1 txpkts work requests"); SYSCTL_ADD_COUNTER_U64(&pi->ctx, children, OID_AUTO, "r_enqueues", CTLFLAG_RD, &txq->r->enqueues, "# of enqueues to the mp_ring for this queue"); SYSCTL_ADD_COUNTER_U64(&pi->ctx, children, OID_AUTO, "r_drops", CTLFLAG_RD, &txq->r->drops, "# of drops in the mp_ring for this queue"); SYSCTL_ADD_COUNTER_U64(&pi->ctx, children, OID_AUTO, "r_starts", CTLFLAG_RD, &txq->r->starts, "# of normal consumer starts in the mp_ring for this queue"); SYSCTL_ADD_COUNTER_U64(&pi->ctx, children, OID_AUTO, "r_stalls", CTLFLAG_RD, &txq->r->stalls, "# of consumer stalls in the mp_ring for this queue"); SYSCTL_ADD_COUNTER_U64(&pi->ctx, children, OID_AUTO, "r_restarts", CTLFLAG_RD, &txq->r->restarts, "# of consumer restarts in the mp_ring for this queue"); SYSCTL_ADD_COUNTER_U64(&pi->ctx, children, OID_AUTO, "r_abdications", CTLFLAG_RD, &txq->r->abdications, "# of consumer abdications in the mp_ring for this queue"); return (0); } static int free_txq(struct port_info *pi, struct sge_txq *txq) { int rc; struct adapter *sc = pi->adapter; struct sge_eq *eq = &txq->eq; rc = free_eq(sc, eq); if (rc) return (rc); sglist_free(txq->gl); free(txq->sdesc, M_CXGBE); mp_ring_free(txq->r); bzero(txq, sizeof(*txq)); return (0); } static void oneseg_dma_callback(void *arg, bus_dma_segment_t *segs, int nseg, int error) { bus_addr_t *ba = arg; KASSERT(nseg == 1, ("%s meant for single segment mappings only.", __func__)); *ba = error ? 0 : segs->ds_addr; } static inline void ring_fl_db(struct adapter *sc, struct sge_fl *fl) { uint32_t n, v; n = IDXDIFF(fl->pidx / 8, fl->dbidx, fl->sidx); MPASS(n > 0); wmb(); v = fl->dbval | V_PIDX(n); if (fl->udb) *fl->udb = htole32(v); else t4_write_reg(sc, MYPF_REG(A_SGE_PF_KDOORBELL), v); IDXINCR(fl->dbidx, n, fl->sidx); } /* * Fills up the freelist by allocating upto 'n' buffers. Buffers that are * recycled do not count towards this allocation budget. * * Returns non-zero to indicate that this freelist should be added to the list * of starving freelists. */ static int refill_fl(struct adapter *sc, struct sge_fl *fl, int n) { __be64 *d; struct fl_sdesc *sd; uintptr_t pa; caddr_t cl; struct cluster_layout *cll; struct sw_zone_info *swz; struct cluster_metadata *clm; uint16_t max_pidx; uint16_t hw_cidx = fl->hw_cidx; /* stable snapshot */ FL_LOCK_ASSERT_OWNED(fl); /* * We always stop at the begining of the hardware descriptor that's just * before the one with the hw cidx. This is to avoid hw pidx = hw cidx, * which would mean an empty freelist to the chip. */ max_pidx = __predict_false(hw_cidx == 0) ? fl->sidx - 1 : hw_cidx - 1; if (fl->pidx == max_pidx * 8) return (0); d = &fl->desc[fl->pidx]; sd = &fl->sdesc[fl->pidx]; cll = &fl->cll_def; /* default layout */ swz = &sc->sge.sw_zone_info[cll->zidx]; while (n > 0) { if (sd->cl != NULL) { if (sd->nmbuf == 0) { /* * Fast recycle without involving any atomics on * the cluster's metadata (if the cluster has * metadata). This happens when all frames * received in the cluster were small enough to * fit within a single mbuf each. */ fl->cl_fast_recycled++; #ifdef INVARIANTS clm = cl_metadata(sc, fl, &sd->cll, sd->cl); if (clm != NULL) MPASS(clm->refcount == 1); #endif goto recycled_fast; } /* * Cluster is guaranteed to have metadata. Clusters * without metadata always take the fast recycle path * when they're recycled. */ clm = cl_metadata(sc, fl, &sd->cll, sd->cl); MPASS(clm != NULL); if (atomic_fetchadd_int(&clm->refcount, -1) == 1) { fl->cl_recycled++; counter_u64_add(extfree_rels, 1); goto recycled; } sd->cl = NULL; /* gave up my reference */ } MPASS(sd->cl == NULL); alloc: cl = uma_zalloc(swz->zone, M_NOWAIT); if (__predict_false(cl == NULL)) { if (cll == &fl->cll_alt || fl->cll_alt.zidx == -1 || fl->cll_def.zidx == fl->cll_alt.zidx) break; /* fall back to the safe zone */ cll = &fl->cll_alt; swz = &sc->sge.sw_zone_info[cll->zidx]; goto alloc; } fl->cl_allocated++; n--; pa = pmap_kextract((vm_offset_t)cl); pa += cll->region1; sd->cl = cl; sd->cll = *cll; *d = htobe64(pa | cll->hwidx); clm = cl_metadata(sc, fl, cll, cl); if (clm != NULL) { recycled: #ifdef INVARIANTS clm->sd = sd; #endif clm->refcount = 1; } sd->nmbuf = 0; recycled_fast: d++; sd++; if (__predict_false(++fl->pidx % 8 == 0)) { uint16_t pidx = fl->pidx / 8; if (__predict_false(pidx == fl->sidx)) { fl->pidx = 0; pidx = 0; sd = fl->sdesc; d = fl->desc; } if (pidx == max_pidx) break; if (IDXDIFF(pidx, fl->dbidx, fl->sidx) >= 4) ring_fl_db(sc, fl); } } if (fl->pidx / 8 != fl->dbidx) ring_fl_db(sc, fl); return (FL_RUNNING_LOW(fl) && !(fl->flags & FL_STARVING)); } /* * Attempt to refill all starving freelists. */ static void refill_sfl(void *arg) { struct adapter *sc = arg; struct sge_fl *fl, *fl_temp; mtx_lock(&sc->sfl_lock); TAILQ_FOREACH_SAFE(fl, &sc->sfl, link, fl_temp) { FL_LOCK(fl); refill_fl(sc, fl, 64); if (FL_NOT_RUNNING_LOW(fl) || fl->flags & FL_DOOMED) { TAILQ_REMOVE(&sc->sfl, fl, link); fl->flags &= ~FL_STARVING; } FL_UNLOCK(fl); } if (!TAILQ_EMPTY(&sc->sfl)) callout_schedule(&sc->sfl_callout, hz / 5); mtx_unlock(&sc->sfl_lock); } static int alloc_fl_sdesc(struct sge_fl *fl) { fl->sdesc = malloc(fl->sidx * 8 * sizeof(struct fl_sdesc), M_CXGBE, M_ZERO | M_WAITOK); return (0); } static void free_fl_sdesc(struct adapter *sc, struct sge_fl *fl) { struct fl_sdesc *sd; struct cluster_metadata *clm; struct cluster_layout *cll; int i; sd = fl->sdesc; for (i = 0; i < fl->sidx * 8; i++, sd++) { if (sd->cl == NULL) continue; cll = &sd->cll; clm = cl_metadata(sc, fl, cll, sd->cl); if (sd->nmbuf == 0) uma_zfree(sc->sge.sw_zone_info[cll->zidx].zone, sd->cl); else if (clm && atomic_fetchadd_int(&clm->refcount, -1) == 1) { uma_zfree(sc->sge.sw_zone_info[cll->zidx].zone, sd->cl); counter_u64_add(extfree_rels, 1); } sd->cl = NULL; } free(fl->sdesc, M_CXGBE); fl->sdesc = NULL; } static inline void get_pkt_gl(struct mbuf *m, struct sglist *gl) { int rc; M_ASSERTPKTHDR(m); sglist_reset(gl); rc = sglist_append_mbuf(gl, m); if (__predict_false(rc != 0)) { panic("%s: mbuf %p (%d segs) was vetted earlier but now fails " "with %d.", __func__, m, mbuf_nsegs(m), rc); } KASSERT(gl->sg_nseg == mbuf_nsegs(m), ("%s: nsegs changed for mbuf %p from %d to %d", __func__, m, mbuf_nsegs(m), gl->sg_nseg)); KASSERT(gl->sg_nseg > 0 && gl->sg_nseg <= (needs_tso(m) ? TX_SGL_SEGS_TSO : TX_SGL_SEGS), ("%s: %d segments, should have been 1 <= nsegs <= %d", __func__, gl->sg_nseg, needs_tso(m) ? TX_SGL_SEGS_TSO : TX_SGL_SEGS)); } /* * len16 for a txpkt WR with a GL. Includes the firmware work request header. */ static inline u_int txpkt_len16(u_int nsegs, u_int tso) { u_int n; MPASS(nsegs > 0); nsegs--; /* first segment is part of ulptx_sgl */ n = sizeof(struct fw_eth_tx_pkt_wr) + sizeof(struct cpl_tx_pkt_core) + sizeof(struct ulptx_sgl) + 8 * ((3 * nsegs) / 2 + (nsegs & 1)); if (tso) n += sizeof(struct cpl_tx_pkt_lso_core); return (howmany(n, 16)); } /* * len16 for a txpkts type 0 WR with a GL. Does not include the firmware work * request header. */ static inline u_int txpkts0_len16(u_int nsegs) { u_int n; MPASS(nsegs > 0); nsegs--; /* first segment is part of ulptx_sgl */ n = sizeof(struct ulp_txpkt) + sizeof(struct ulptx_idata) + sizeof(struct cpl_tx_pkt_core) + sizeof(struct ulptx_sgl) + 8 * ((3 * nsegs) / 2 + (nsegs & 1)); return (howmany(n, 16)); } /* * len16 for a txpkts type 1 WR with a GL. Does not include the firmware work * request header. */ static inline u_int txpkts1_len16(void) { u_int n; n = sizeof(struct cpl_tx_pkt_core) + sizeof(struct ulptx_sgl); return (howmany(n, 16)); } static inline u_int imm_payload(u_int ndesc) { u_int n; n = ndesc * EQ_ESIZE - sizeof(struct fw_eth_tx_pkt_wr) - sizeof(struct cpl_tx_pkt_core); return (n); } /* * Write a txpkt WR for this packet to the hardware descriptors, update the * software descriptor, and advance the pidx. It is guaranteed that enough * descriptors are available. * * The return value is the # of hardware descriptors used. */ static u_int write_txpkt_wr(struct sge_txq *txq, struct fw_eth_tx_pkt_wr *wr, struct mbuf *m0, u_int available) { struct sge_eq *eq = &txq->eq; struct tx_sdesc *txsd; struct cpl_tx_pkt_core *cpl; uint32_t ctrl; /* used in many unrelated places */ uint64_t ctrl1; int len16, ndesc, pktlen, nsegs; caddr_t dst; TXQ_LOCK_ASSERT_OWNED(txq); M_ASSERTPKTHDR(m0); MPASS(available > 0 && available < eq->sidx); len16 = mbuf_len16(m0); nsegs = mbuf_nsegs(m0); pktlen = m0->m_pkthdr.len; ctrl = sizeof(struct cpl_tx_pkt_core); if (needs_tso(m0)) ctrl += sizeof(struct cpl_tx_pkt_lso_core); else if (pktlen <= imm_payload(2) && available >= 2) { /* Immediate data. Recalculate len16 and set nsegs to 0. */ ctrl += pktlen; len16 = howmany(sizeof(struct fw_eth_tx_pkt_wr) + sizeof(struct cpl_tx_pkt_core) + pktlen, 16); nsegs = 0; } ndesc = howmany(len16, EQ_ESIZE / 16); MPASS(ndesc <= available); /* Firmware work request header */ MPASS(wr == (void *)&eq->desc[eq->pidx]); wr->op_immdlen = htobe32(V_FW_WR_OP(FW_ETH_TX_PKT_WR) | V_FW_ETH_TX_PKT_WR_IMMDLEN(ctrl)); ctrl = V_FW_WR_LEN16(len16); wr->equiq_to_len16 = htobe32(ctrl); wr->r3 = 0; if (needs_tso(m0)) { struct cpl_tx_pkt_lso_core *lso = (void *)(wr + 1); KASSERT(m0->m_pkthdr.l2hlen > 0 && m0->m_pkthdr.l3hlen > 0 && m0->m_pkthdr.l4hlen > 0, ("%s: mbuf %p needs TSO but missing header lengths", __func__, m0)); ctrl = V_LSO_OPCODE(CPL_TX_PKT_LSO) | F_LSO_FIRST_SLICE | F_LSO_LAST_SLICE | V_LSO_IPHDR_LEN(m0->m_pkthdr.l3hlen >> 2) | V_LSO_TCPHDR_LEN(m0->m_pkthdr.l4hlen >> 2); if (m0->m_pkthdr.l2hlen == sizeof(struct ether_vlan_header)) ctrl |= V_LSO_ETHHDR_LEN(1); if (m0->m_pkthdr.l3hlen == sizeof(struct ip6_hdr)) ctrl |= F_LSO_IPV6; lso->lso_ctrl = htobe32(ctrl); lso->ipid_ofst = htobe16(0); lso->mss = htobe16(m0->m_pkthdr.tso_segsz); lso->seqno_offset = htobe32(0); lso->len = htobe32(pktlen); cpl = (void *)(lso + 1); txq->tso_wrs++; } else cpl = (void *)(wr + 1); /* Checksum offload */ ctrl1 = 0; if (needs_l3_csum(m0) == 0) ctrl1 |= F_TXPKT_IPCSUM_DIS; if (needs_l4_csum(m0) == 0) ctrl1 |= F_TXPKT_L4CSUM_DIS; if (m0->m_pkthdr.csum_flags & (CSUM_IP | CSUM_TCP | CSUM_UDP | CSUM_UDP_IPV6 | CSUM_TCP_IPV6 | CSUM_TSO)) txq->txcsum++; /* some hardware assistance provided */ /* VLAN tag insertion */ if (needs_vlan_insertion(m0)) { ctrl1 |= F_TXPKT_VLAN_VLD | V_TXPKT_VLAN(m0->m_pkthdr.ether_vtag); txq->vlan_insertion++; } /* CPL header */ cpl->ctrl0 = txq->cpl_ctrl0; cpl->pack = 0; cpl->len = htobe16(pktlen); cpl->ctrl1 = htobe64(ctrl1); /* SGL */ dst = (void *)(cpl + 1); if (nsegs > 0) { write_gl_to_txd(txq, m0, &dst, eq->sidx - ndesc < eq->pidx); txq->sgl_wrs++; } else { struct mbuf *m; for (m = m0; m != NULL; m = m->m_next) { copy_to_txd(eq, mtod(m, caddr_t), &dst, m->m_len); #ifdef INVARIANTS pktlen -= m->m_len; #endif } #ifdef INVARIANTS KASSERT(pktlen == 0, ("%s: %d bytes left.", __func__, pktlen)); #endif txq->imm_wrs++; } txq->txpkt_wrs++; txsd = &txq->sdesc[eq->pidx]; txsd->m = m0; txsd->desc_used = ndesc; return (ndesc); } static int try_txpkts(struct mbuf *m, struct mbuf *n, struct txpkts *txp, u_int available) { u_int needed, nsegs1, nsegs2, l1, l2; if (cannot_use_txpkts(m) || cannot_use_txpkts(n)) return (1); nsegs1 = mbuf_nsegs(m); nsegs2 = mbuf_nsegs(n); if (nsegs1 + nsegs2 == 2) { txp->wr_type = 1; l1 = l2 = txpkts1_len16(); } else { txp->wr_type = 0; l1 = txpkts0_len16(nsegs1); l2 = txpkts0_len16(nsegs2); } txp->len16 = howmany(sizeof(struct fw_eth_tx_pkts_wr), 16) + l1 + l2; needed = howmany(txp->len16, EQ_ESIZE / 16); if (needed > SGE_MAX_WR_NDESC || needed > available) return (1); txp->plen = m->m_pkthdr.len + n->m_pkthdr.len; if (txp->plen > 65535) return (1); txp->npkt = 2; set_mbuf_len16(m, l1); set_mbuf_len16(n, l2); return (0); } static int add_to_txpkts(struct mbuf *m, struct txpkts *txp, u_int available) { u_int plen, len16, needed, nsegs; MPASS(txp->wr_type == 0 || txp->wr_type == 1); nsegs = mbuf_nsegs(m); if (needs_tso(m) || (txp->wr_type == 1 && nsegs != 1)) return (1); plen = txp->plen + m->m_pkthdr.len; if (plen > 65535) return (1); if (txp->wr_type == 0) len16 = txpkts0_len16(nsegs); else len16 = txpkts1_len16(); needed = howmany(txp->len16 + len16, EQ_ESIZE / 16); if (needed > SGE_MAX_WR_NDESC || needed > available) return (1); txp->npkt++; txp->plen = plen; txp->len16 += len16; set_mbuf_len16(m, len16); return (0); } /* * Write a txpkts WR for the packets in txp to the hardware descriptors, update * the software descriptor, and advance the pidx. It is guaranteed that enough * descriptors are available. * * The return value is the # of hardware descriptors used. */ static u_int write_txpkts_wr(struct sge_txq *txq, struct fw_eth_tx_pkts_wr *wr, struct mbuf *m0, const struct txpkts *txp, u_int available) { struct sge_eq *eq = &txq->eq; struct tx_sdesc *txsd; struct cpl_tx_pkt_core *cpl; uint32_t ctrl; uint64_t ctrl1; int ndesc, checkwrap; struct mbuf *m; void *flitp; TXQ_LOCK_ASSERT_OWNED(txq); MPASS(txp->npkt > 0); MPASS(txp->plen < 65536); MPASS(m0 != NULL); MPASS(m0->m_nextpkt != NULL); MPASS(txp->len16 <= howmany(SGE_MAX_WR_LEN, 16)); MPASS(available > 0 && available < eq->sidx); ndesc = howmany(txp->len16, EQ_ESIZE / 16); MPASS(ndesc <= available); MPASS(wr == (void *)&eq->desc[eq->pidx]); wr->op_pkd = htobe32(V_FW_WR_OP(FW_ETH_TX_PKTS_WR)); ctrl = V_FW_WR_LEN16(txp->len16); wr->equiq_to_len16 = htobe32(ctrl); wr->plen = htobe16(txp->plen); wr->npkt = txp->npkt; wr->r3 = 0; wr->type = txp->wr_type; flitp = wr + 1; /* * At this point we are 16B into a hardware descriptor. If checkwrap is * set then we know the WR is going to wrap around somewhere. We'll * check for that at appropriate points. */ checkwrap = eq->sidx - ndesc < eq->pidx; for (m = m0; m != NULL; m = m->m_nextpkt) { if (txp->wr_type == 0) { struct ulp_txpkt *ulpmc; struct ulptx_idata *ulpsc; /* ULP master command */ ulpmc = flitp; ulpmc->cmd_dest = htobe32(V_ULPTX_CMD(ULP_TX_PKT) | V_ULP_TXPKT_DEST(0) | V_ULP_TXPKT_FID(eq->iqid)); ulpmc->len = htobe32(mbuf_len16(m)); /* ULP subcommand */ ulpsc = (void *)(ulpmc + 1); ulpsc->cmd_more = htobe32(V_ULPTX_CMD(ULP_TX_SC_IMM) | F_ULP_TX_SC_MORE); ulpsc->len = htobe32(sizeof(struct cpl_tx_pkt_core)); cpl = (void *)(ulpsc + 1); if (checkwrap && (uintptr_t)cpl == (uintptr_t)&eq->desc[eq->sidx]) cpl = (void *)&eq->desc[0]; txq->txpkts0_pkts += txp->npkt; txq->txpkts0_wrs++; } else { cpl = flitp; txq->txpkts1_pkts += txp->npkt; txq->txpkts1_wrs++; } /* Checksum offload */ ctrl1 = 0; if (needs_l3_csum(m) == 0) ctrl1 |= F_TXPKT_IPCSUM_DIS; if (needs_l4_csum(m) == 0) ctrl1 |= F_TXPKT_L4CSUM_DIS; if (m->m_pkthdr.csum_flags & (CSUM_IP | CSUM_TCP | CSUM_UDP | CSUM_UDP_IPV6 | CSUM_TCP_IPV6 | CSUM_TSO)) txq->txcsum++; /* some hardware assistance provided */ /* VLAN tag insertion */ if (needs_vlan_insertion(m)) { ctrl1 |= F_TXPKT_VLAN_VLD | V_TXPKT_VLAN(m->m_pkthdr.ether_vtag); txq->vlan_insertion++; } /* CPL header */ cpl->ctrl0 = txq->cpl_ctrl0; cpl->pack = 0; cpl->len = htobe16(m->m_pkthdr.len); cpl->ctrl1 = htobe64(ctrl1); flitp = cpl + 1; if (checkwrap && (uintptr_t)flitp == (uintptr_t)&eq->desc[eq->sidx]) flitp = (void *)&eq->desc[0]; write_gl_to_txd(txq, m, (caddr_t *)(&flitp), checkwrap); } txsd = &txq->sdesc[eq->pidx]; txsd->m = m0; txsd->desc_used = ndesc; return (ndesc); } /* * If the SGL ends on an address that is not 16 byte aligned, this function will * add a 0 filled flit at the end. */ static void write_gl_to_txd(struct sge_txq *txq, struct mbuf *m, caddr_t *to, int checkwrap) { struct sge_eq *eq = &txq->eq; struct sglist *gl = txq->gl; struct sglist_seg *seg; __be64 *flitp, *wrap; struct ulptx_sgl *usgl; int i, nflits, nsegs; KASSERT(((uintptr_t)(*to) & 0xf) == 0, ("%s: SGL must start at a 16 byte boundary: %p", __func__, *to)); MPASS((uintptr_t)(*to) >= (uintptr_t)&eq->desc[0]); MPASS((uintptr_t)(*to) < (uintptr_t)&eq->desc[eq->sidx]); get_pkt_gl(m, gl); nsegs = gl->sg_nseg; MPASS(nsegs > 0); nflits = (3 * (nsegs - 1)) / 2 + ((nsegs - 1) & 1) + 2; flitp = (__be64 *)(*to); wrap = (__be64 *)(&eq->desc[eq->sidx]); seg = &gl->sg_segs[0]; usgl = (void *)flitp; /* * We start at a 16 byte boundary somewhere inside the tx descriptor * ring, so we're at least 16 bytes away from the status page. There is * no chance of a wrap around in the middle of usgl (which is 16 bytes). */ usgl->cmd_nsge = htobe32(V_ULPTX_CMD(ULP_TX_SC_DSGL) | V_ULPTX_NSGE(nsegs)); usgl->len0 = htobe32(seg->ss_len); usgl->addr0 = htobe64(seg->ss_paddr); seg++; if (checkwrap == 0 || (uintptr_t)(flitp + nflits) <= (uintptr_t)wrap) { /* Won't wrap around at all */ for (i = 0; i < nsegs - 1; i++, seg++) { usgl->sge[i / 2].len[i & 1] = htobe32(seg->ss_len); usgl->sge[i / 2].addr[i & 1] = htobe64(seg->ss_paddr); } if (i & 1) usgl->sge[i / 2].len[1] = htobe32(0); flitp += nflits; } else { /* Will wrap somewhere in the rest of the SGL */ /* 2 flits already written, write the rest flit by flit */ flitp = (void *)(usgl + 1); for (i = 0; i < nflits - 2; i++) { if (flitp == wrap) flitp = (void *)eq->desc; *flitp++ = get_flit(seg, nsegs - 1, i); } } if (nflits & 1) { MPASS(((uintptr_t)flitp) & 0xf); *flitp++ = 0; } MPASS((((uintptr_t)flitp) & 0xf) == 0); if (__predict_false(flitp == wrap)) *to = (void *)eq->desc; else *to = (void *)flitp; } static inline void copy_to_txd(struct sge_eq *eq, caddr_t from, caddr_t *to, int len) { MPASS((uintptr_t)(*to) >= (uintptr_t)&eq->desc[0]); MPASS((uintptr_t)(*to) < (uintptr_t)&eq->desc[eq->sidx]); if (__predict_true((uintptr_t)(*to) + len <= (uintptr_t)&eq->desc[eq->sidx])) { bcopy(from, *to, len); (*to) += len; } else { int portion = (uintptr_t)&eq->desc[eq->sidx] - (uintptr_t)(*to); bcopy(from, *to, portion); from += portion; portion = len - portion; /* remaining */ bcopy(from, (void *)eq->desc, portion); (*to) = (caddr_t)eq->desc + portion; } } static inline void ring_eq_db(struct adapter *sc, struct sge_eq *eq, u_int n) { u_int db; MPASS(n > 0); db = eq->doorbells; if (n > 1) clrbit(&db, DOORBELL_WCWR); wmb(); switch (ffs(db) - 1) { case DOORBELL_UDB: *eq->udb = htole32(V_QID(eq->udb_qid) | V_PIDX(n)); break; case DOORBELL_WCWR: { volatile uint64_t *dst, *src; int i; /* * Queues whose 128B doorbell segment fits in the page do not * use relative qid (udb_qid is always 0). Only queues with * doorbell segments can do WCWR. */ KASSERT(eq->udb_qid == 0 && n == 1, ("%s: inappropriate doorbell (0x%x, %d, %d) for eq %p", __func__, eq->doorbells, n, eq->dbidx, eq)); dst = (volatile void *)((uintptr_t)eq->udb + UDBS_WR_OFFSET - UDBS_DB_OFFSET); i = eq->dbidx; src = (void *)&eq->desc[i]; while (src != (void *)&eq->desc[i + 1]) *dst++ = *src++; wmb(); break; } case DOORBELL_UDBWC: *eq->udb = htole32(V_QID(eq->udb_qid) | V_PIDX(n)); wmb(); break; case DOORBELL_KDB: t4_write_reg(sc, MYPF_REG(A_SGE_PF_KDOORBELL), V_QID(eq->cntxt_id) | V_PIDX(n)); break; } IDXINCR(eq->dbidx, n, eq->sidx); } static inline u_int reclaimable_tx_desc(struct sge_eq *eq) { uint16_t hw_cidx; hw_cidx = read_hw_cidx(eq); return (IDXDIFF(hw_cidx, eq->cidx, eq->sidx)); } static inline u_int total_available_tx_desc(struct sge_eq *eq) { uint16_t hw_cidx, pidx; hw_cidx = read_hw_cidx(eq); pidx = eq->pidx; if (pidx == hw_cidx) return (eq->sidx - 1); else return (IDXDIFF(hw_cidx, pidx, eq->sidx) - 1); } static inline uint16_t read_hw_cidx(struct sge_eq *eq) { struct sge_qstat *spg = (void *)&eq->desc[eq->sidx]; uint16_t cidx = spg->cidx; /* stable snapshot */ return (be16toh(cidx)); } /* * Reclaim 'n' descriptors approximately. */ static u_int reclaim_tx_descs(struct sge_txq *txq, u_int n) { struct tx_sdesc *txsd; struct sge_eq *eq = &txq->eq; u_int can_reclaim, reclaimed; TXQ_LOCK_ASSERT_OWNED(txq); MPASS(n > 0); reclaimed = 0; can_reclaim = reclaimable_tx_desc(eq); while (can_reclaim && reclaimed < n) { int ndesc; struct mbuf *m, *nextpkt; txsd = &txq->sdesc[eq->cidx]; ndesc = txsd->desc_used; /* Firmware doesn't return "partial" credits. */ KASSERT(can_reclaim >= ndesc, ("%s: unexpected number of credits: %d, %d", __func__, can_reclaim, ndesc)); for (m = txsd->m; m != NULL; m = nextpkt) { nextpkt = m->m_nextpkt; m->m_nextpkt = NULL; m_freem(m); } reclaimed += ndesc; can_reclaim -= ndesc; IDXINCR(eq->cidx, ndesc, eq->sidx); } return (reclaimed); } static void tx_reclaim(void *arg, int n) { struct sge_txq *txq = arg; struct sge_eq *eq = &txq->eq; do { if (TXQ_TRYLOCK(txq) == 0) break; n = reclaim_tx_descs(txq, 32); if (eq->cidx == eq->pidx) eq->equeqidx = eq->pidx; TXQ_UNLOCK(txq); } while (n > 0); } static __be64 get_flit(struct sglist_seg *segs, int nsegs, int idx) { int i = (idx / 3) * 2; switch (idx % 3) { case 0: { __be64 rc; rc = htobe32(segs[i].ss_len); if (i + 1 < nsegs) rc |= (uint64_t)htobe32(segs[i + 1].ss_len) << 32; return (rc); } case 1: return (htobe64(segs[i].ss_paddr)); case 2: return (htobe64(segs[i + 1].ss_paddr)); } return (0); } static void find_best_refill_source(struct adapter *sc, struct sge_fl *fl, int maxp) { int8_t zidx, hwidx, idx; uint16_t region1, region3; int spare, spare_needed, n; struct sw_zone_info *swz; struct hw_buf_info *hwb, *hwb_list = &sc->sge.hw_buf_info[0]; /* * Buffer Packing: Look for PAGE_SIZE or larger zone which has a bufsize * large enough for the max payload and cluster metadata. Otherwise * settle for the largest bufsize that leaves enough room in the cluster * for metadata. * * Without buffer packing: Look for the smallest zone which has a * bufsize large enough for the max payload. Settle for the largest * bufsize available if there's nothing big enough for max payload. */ spare_needed = fl->flags & FL_BUF_PACKING ? CL_METADATA_SIZE : 0; swz = &sc->sge.sw_zone_info[0]; hwidx = -1; for (zidx = 0; zidx < SW_ZONE_SIZES; zidx++, swz++) { if (swz->size > largest_rx_cluster) { if (__predict_true(hwidx != -1)) break; /* * This is a misconfiguration. largest_rx_cluster is * preventing us from finding a refill source. See * dev.t5nex..buffer_sizes to figure out why. */ device_printf(sc->dev, "largest_rx_cluster=%u leaves no" " refill source for fl %p (dma %u). Ignored.\n", largest_rx_cluster, fl, maxp); } for (idx = swz->head_hwidx; idx != -1; idx = hwb->next) { hwb = &hwb_list[idx]; spare = swz->size - hwb->size; if (spare < spare_needed) continue; hwidx = idx; /* best option so far */ if (hwb->size >= maxp) { if ((fl->flags & FL_BUF_PACKING) == 0) goto done; /* stop looking (not packing) */ if (swz->size >= safest_rx_cluster) goto done; /* stop looking (packing) */ } break; /* keep looking, next zone */ } } done: /* A usable hwidx has been located. */ MPASS(hwidx != -1); hwb = &hwb_list[hwidx]; zidx = hwb->zidx; swz = &sc->sge.sw_zone_info[zidx]; region1 = 0; region3 = swz->size - hwb->size; /* * Stay within this zone and see if there is a better match when mbuf * inlining is allowed. Remember that the hwidx's are sorted in * decreasing order of size (so in increasing order of spare area). */ for (idx = hwidx; idx != -1; idx = hwb->next) { hwb = &hwb_list[idx]; spare = swz->size - hwb->size; if (allow_mbufs_in_cluster == 0 || hwb->size < maxp) break; /* * Do not inline mbufs if doing so would violate the pad/pack * boundary alignment requirement. */ if (fl_pad && (MSIZE % sc->sge.pad_boundary) != 0) continue; if (fl->flags & FL_BUF_PACKING && (MSIZE % sc->sge.pack_boundary) != 0) continue; if (spare < CL_METADATA_SIZE + MSIZE) continue; n = (spare - CL_METADATA_SIZE) / MSIZE; if (n > howmany(hwb->size, maxp)) break; hwidx = idx; if (fl->flags & FL_BUF_PACKING) { region1 = n * MSIZE; region3 = spare - region1; } else { region1 = MSIZE; region3 = spare - region1; break; } } KASSERT(zidx >= 0 && zidx < SW_ZONE_SIZES, ("%s: bad zone %d for fl %p, maxp %d", __func__, zidx, fl, maxp)); KASSERT(hwidx >= 0 && hwidx <= SGE_FLBUF_SIZES, ("%s: bad hwidx %d for fl %p, maxp %d", __func__, hwidx, fl, maxp)); KASSERT(region1 + sc->sge.hw_buf_info[hwidx].size + region3 == sc->sge.sw_zone_info[zidx].size, ("%s: bad buffer layout for fl %p, maxp %d. " "cl %d; r1 %d, payload %d, r3 %d", __func__, fl, maxp, sc->sge.sw_zone_info[zidx].size, region1, sc->sge.hw_buf_info[hwidx].size, region3)); if (fl->flags & FL_BUF_PACKING || region1 > 0) { KASSERT(region3 >= CL_METADATA_SIZE, ("%s: no room for metadata. fl %p, maxp %d; " "cl %d; r1 %d, payload %d, r3 %d", __func__, fl, maxp, sc->sge.sw_zone_info[zidx].size, region1, sc->sge.hw_buf_info[hwidx].size, region3)); KASSERT(region1 % MSIZE == 0, ("%s: bad mbuf region for fl %p, maxp %d. " "cl %d; r1 %d, payload %d, r3 %d", __func__, fl, maxp, sc->sge.sw_zone_info[zidx].size, region1, sc->sge.hw_buf_info[hwidx].size, region3)); } fl->cll_def.zidx = zidx; fl->cll_def.hwidx = hwidx; fl->cll_def.region1 = region1; fl->cll_def.region3 = region3; } static void find_safe_refill_source(struct adapter *sc, struct sge_fl *fl) { struct sge *s = &sc->sge; struct hw_buf_info *hwb; struct sw_zone_info *swz; int spare; int8_t hwidx; if (fl->flags & FL_BUF_PACKING) hwidx = s->safe_hwidx2; /* with room for metadata */ else if (allow_mbufs_in_cluster && s->safe_hwidx2 != -1) { hwidx = s->safe_hwidx2; hwb = &s->hw_buf_info[hwidx]; swz = &s->sw_zone_info[hwb->zidx]; spare = swz->size - hwb->size; /* no good if there isn't room for an mbuf as well */ if (spare < CL_METADATA_SIZE + MSIZE) hwidx = s->safe_hwidx1; } else hwidx = s->safe_hwidx1; if (hwidx == -1) { /* No fallback source */ fl->cll_alt.hwidx = -1; fl->cll_alt.zidx = -1; return; } hwb = &s->hw_buf_info[hwidx]; swz = &s->sw_zone_info[hwb->zidx]; spare = swz->size - hwb->size; fl->cll_alt.hwidx = hwidx; fl->cll_alt.zidx = hwb->zidx; if (allow_mbufs_in_cluster && (fl_pad == 0 || (MSIZE % sc->sge.pad_boundary) == 0)) fl->cll_alt.region1 = ((spare - CL_METADATA_SIZE) / MSIZE) * MSIZE; else fl->cll_alt.region1 = 0; fl->cll_alt.region3 = spare - fl->cll_alt.region1; } static void add_fl_to_sfl(struct adapter *sc, struct sge_fl *fl) { mtx_lock(&sc->sfl_lock); FL_LOCK(fl); if ((fl->flags & FL_DOOMED) == 0) { fl->flags |= FL_STARVING; TAILQ_INSERT_TAIL(&sc->sfl, fl, link); callout_reset(&sc->sfl_callout, hz / 5, refill_sfl, sc); } FL_UNLOCK(fl); mtx_unlock(&sc->sfl_lock); } static void handle_wrq_egr_update(struct adapter *sc, struct sge_eq *eq) { struct sge_wrq *wrq = (void *)eq; atomic_readandclear_int(&eq->equiq); taskqueue_enqueue(sc->tq[eq->tx_chan], &wrq->wrq_tx_task); } static void handle_eth_egr_update(struct adapter *sc, struct sge_eq *eq) { struct sge_txq *txq = (void *)eq; MPASS((eq->flags & EQ_TYPEMASK) == EQ_ETH); atomic_readandclear_int(&eq->equiq); mp_ring_check_drainage(txq->r, 0); taskqueue_enqueue(sc->tq[eq->tx_chan], &txq->tx_reclaim_task); } static int handle_sge_egr_update(struct sge_iq *iq, const struct rss_header *rss, struct mbuf *m) { const struct cpl_sge_egr_update *cpl = (const void *)(rss + 1); unsigned int qid = G_EGR_QID(ntohl(cpl->opcode_qid)); struct adapter *sc = iq->adapter; struct sge *s = &sc->sge; struct sge_eq *eq; static void (*h[])(struct adapter *, struct sge_eq *) = {NULL, &handle_wrq_egr_update, &handle_eth_egr_update, &handle_wrq_egr_update}; KASSERT(m == NULL, ("%s: payload with opcode %02x", __func__, rss->opcode)); eq = s->eqmap[qid - s->eq_start]; (*h[eq->flags & EQ_TYPEMASK])(sc, eq); return (0); } /* handle_fw_msg works for both fw4_msg and fw6_msg because this is valid */ CTASSERT(offsetof(struct cpl_fw4_msg, data) == \ offsetof(struct cpl_fw6_msg, data)); static int handle_fw_msg(struct sge_iq *iq, const struct rss_header *rss, struct mbuf *m) { struct adapter *sc = iq->adapter; const struct cpl_fw6_msg *cpl = (const void *)(rss + 1); KASSERT(m == NULL, ("%s: payload with opcode %02x", __func__, rss->opcode)); if (cpl->type == FW_TYPE_RSSCPL || cpl->type == FW6_TYPE_RSSCPL) { const struct rss_header *rss2; rss2 = (const struct rss_header *)&cpl->data[0]; return (sc->cpl_handler[rss2->opcode](iq, rss2, m)); } return (sc->fw_msg_handler[cpl->type](sc, &cpl->data[0])); } static int sysctl_uint16(SYSCTL_HANDLER_ARGS) { uint16_t *id = arg1; int i = *id; return sysctl_handle_int(oidp, &i, 0, req); } static int sysctl_bufsizes(SYSCTL_HANDLER_ARGS) { struct sge *s = arg1; struct hw_buf_info *hwb = &s->hw_buf_info[0]; struct sw_zone_info *swz = &s->sw_zone_info[0]; int i, rc; struct sbuf sb; char c; sbuf_new(&sb, NULL, 32, SBUF_AUTOEXTEND); for (i = 0; i < SGE_FLBUF_SIZES; i++, hwb++) { if (hwb->zidx >= 0 && swz[hwb->zidx].size <= largest_rx_cluster) c = '*'; else c = '\0'; sbuf_printf(&sb, "%u%c ", hwb->size, c); } sbuf_trim(&sb); sbuf_finish(&sb); rc = sysctl_handle_string(oidp, sbuf_data(&sb), sbuf_len(&sb), req); sbuf_delete(&sb); return (rc); } Index: projects/clang-trunk/sys/dev/uart/uart_dev_ns8250.c =================================================================== --- projects/clang-trunk/sys/dev/uart/uart_dev_ns8250.c (revision 283870) +++ projects/clang-trunk/sys/dev/uart/uart_dev_ns8250.c (revision 283871) @@ -1,971 +1,971 @@ /*- * Copyright (c) 2003 Marcel Moolenaar * All rights reserved. * * Redistribution and use in source and binary forms, with or without * modification, are permitted provided that the following conditions * are met: * * 1. Redistributions of source code must retain the above copyright * notice, this list of conditions and the following disclaimer. * 2. Redistributions in binary form must reproduce the above copyright * notice, this list of conditions and the following disclaimer in the * documentation and/or other materials provided with the distribution. * * THIS SOFTWARE IS PROVIDED BY THE AUTHOR ``AS IS'' AND ANY EXPRESS OR * IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES * OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. * IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR ANY DIRECT, INDIRECT, * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT * NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF * THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. */ #include "opt_platform.h" #include __FBSDID("$FreeBSD$"); #include #include #include #include #include #include #include #ifdef FDT #include #include #include #endif #include #include #ifdef FDT #include #endif #include #include #include #include "uart_if.h" #define DEFAULT_RCLK 1843200 static int broken_txfifo = 0; SYSCTL_INT(_hw, OID_AUTO, broken_txfifo, CTLFLAG_RWTUN, &broken_txfifo, 0, "UART FIFO has QEMU emulation bug"); /* * Clear pending interrupts. THRE is cleared by reading IIR. Data * that may have been received gets lost here. */ static void ns8250_clrint(struct uart_bas *bas) { uint8_t iir, lsr; iir = uart_getreg(bas, REG_IIR); while ((iir & IIR_NOPEND) == 0) { iir &= IIR_IMASK; if (iir == IIR_RLS) { lsr = uart_getreg(bas, REG_LSR); if (lsr & (LSR_BI|LSR_FE|LSR_PE)) (void)uart_getreg(bas, REG_DATA); } else if (iir == IIR_RXRDY || iir == IIR_RXTOUT) (void)uart_getreg(bas, REG_DATA); else if (iir == IIR_MLSC) (void)uart_getreg(bas, REG_MSR); uart_barrier(bas); iir = uart_getreg(bas, REG_IIR); } } static int ns8250_delay(struct uart_bas *bas) { int divisor; u_char lcr; lcr = uart_getreg(bas, REG_LCR); uart_setreg(bas, REG_LCR, lcr | LCR_DLAB); uart_barrier(bas); divisor = uart_getreg(bas, REG_DLL) | (uart_getreg(bas, REG_DLH) << 8); uart_barrier(bas); uart_setreg(bas, REG_LCR, lcr); uart_barrier(bas); /* 1/10th the time to transmit 1 character (estimate). */ if (divisor <= 134) return (16000000 * divisor / bas->rclk); return (16000 * divisor / (bas->rclk / 1000)); } static int ns8250_divisor(int rclk, int baudrate) { int actual_baud, divisor; int error; if (baudrate == 0) return (0); divisor = (rclk / (baudrate << 3) + 1) >> 1; if (divisor == 0 || divisor >= 65536) return (0); actual_baud = rclk / (divisor << 4); /* 10 times error in percent: */ error = ((actual_baud - baudrate) * 2000 / baudrate + 1) >> 1; /* 3.0% maximum error tolerance: */ if (error < -30 || error > 30) return (0); return (divisor); } static int ns8250_drain(struct uart_bas *bas, int what) { int delay, limit; delay = ns8250_delay(bas); if (what & UART_DRAIN_TRANSMITTER) { /* * Pick an arbitrary high limit to avoid getting stuck in * an infinite loop when the hardware is broken. Make the * limit high enough to handle large FIFOs. */ limit = 10*1024; while ((uart_getreg(bas, REG_LSR) & LSR_TEMT) == 0 && --limit) DELAY(delay); if (limit == 0) { /* printf("ns8250: transmitter appears stuck... "); */ return (EIO); } } if (what & UART_DRAIN_RECEIVER) { /* * Pick an arbitrary high limit to avoid getting stuck in * an infinite loop when the hardware is broken. Make the * limit high enough to handle large FIFOs and integrated * UARTs. The HP rx2600 for example has 3 UARTs on the * management board that tend to get a lot of data send * to it when the UART is first activated. */ limit=10*4096; while ((uart_getreg(bas, REG_LSR) & LSR_RXRDY) && --limit) { (void)uart_getreg(bas, REG_DATA); uart_barrier(bas); DELAY(delay << 2); } if (limit == 0) { /* printf("ns8250: receiver appears broken... "); */ return (EIO); } } return (0); } /* * We can only flush UARTs with FIFOs. UARTs without FIFOs should be * drained. WARNING: this function clobbers the FIFO setting! */ static void ns8250_flush(struct uart_bas *bas, int what) { uint8_t fcr; fcr = FCR_ENABLE; if (what & UART_FLUSH_TRANSMITTER) fcr |= FCR_XMT_RST; if (what & UART_FLUSH_RECEIVER) fcr |= FCR_RCV_RST; uart_setreg(bas, REG_FCR, fcr); uart_barrier(bas); } static int ns8250_param(struct uart_bas *bas, int baudrate, int databits, int stopbits, int parity) { int divisor; uint8_t lcr; lcr = 0; if (databits >= 8) lcr |= LCR_8BITS; else if (databits == 7) lcr |= LCR_7BITS; else if (databits == 6) lcr |= LCR_6BITS; else lcr |= LCR_5BITS; if (stopbits > 1) lcr |= LCR_STOPB; lcr |= parity << 3; /* Set baudrate. */ if (baudrate > 0) { divisor = ns8250_divisor(bas->rclk, baudrate); if (divisor == 0) return (EINVAL); uart_setreg(bas, REG_LCR, lcr | LCR_DLAB); uart_barrier(bas); uart_setreg(bas, REG_DLL, divisor & 0xff); uart_setreg(bas, REG_DLH, (divisor >> 8) & 0xff); uart_barrier(bas); } /* Set LCR and clear DLAB. */ uart_setreg(bas, REG_LCR, lcr); uart_barrier(bas); return (0); } /* * Low-level UART interface. */ static int ns8250_probe(struct uart_bas *bas); static void ns8250_init(struct uart_bas *bas, int, int, int, int); static void ns8250_term(struct uart_bas *bas); static void ns8250_putc(struct uart_bas *bas, int); static int ns8250_rxready(struct uart_bas *bas); static int ns8250_getc(struct uart_bas *bas, struct mtx *); struct uart_ops uart_ns8250_ops = { .probe = ns8250_probe, .init = ns8250_init, .term = ns8250_term, .putc = ns8250_putc, .rxready = ns8250_rxready, .getc = ns8250_getc, }; static int ns8250_probe(struct uart_bas *bas) { u_char val; /* Check known 0 bits that don't depend on DLAB. */ val = uart_getreg(bas, REG_IIR); if (val & 0x30) return (ENXIO); /* * Bit 6 of the MCR (= 0x40) appears to be 1 for the Sun1699 * chip, but otherwise doesn't seem to have a function. In * other words, uart(4) works regardless. Ignore that bit so * the probe succeeds. */ val = uart_getreg(bas, REG_MCR); if (val & 0xa0) return (ENXIO); return (0); } static void ns8250_init(struct uart_bas *bas, int baudrate, int databits, int stopbits, int parity) { u_char ier; if (bas->rclk == 0) bas->rclk = DEFAULT_RCLK; ns8250_param(bas, baudrate, databits, stopbits, parity); /* Disable all interrupt sources. */ /* * We use 0xe0 instead of 0xf0 as the mask because the XScale PXA * UARTs split the receive time-out interrupt bit out separately as * 0x10. This gets handled by ier_mask and ier_rxbits below. */ ier = uart_getreg(bas, REG_IER) & 0xe0; uart_setreg(bas, REG_IER, ier); uart_barrier(bas); /* Disable the FIFO (if present). */ uart_setreg(bas, REG_FCR, 0); uart_barrier(bas); /* Set RTS & DTR. */ uart_setreg(bas, REG_MCR, MCR_IE | MCR_RTS | MCR_DTR); uart_barrier(bas); ns8250_clrint(bas); } static void ns8250_term(struct uart_bas *bas) { /* Clear RTS & DTR. */ uart_setreg(bas, REG_MCR, MCR_IE); uart_barrier(bas); } static void ns8250_putc(struct uart_bas *bas, int c) { int limit; limit = 250000; while ((uart_getreg(bas, REG_LSR) & LSR_THRE) == 0 && --limit) DELAY(4); uart_setreg(bas, REG_DATA, c); uart_barrier(bas); limit = 250000; while ((uart_getreg(bas, REG_LSR) & LSR_TEMT) == 0 && --limit) DELAY(4); } static int ns8250_rxready(struct uart_bas *bas) { return ((uart_getreg(bas, REG_LSR) & LSR_RXRDY) != 0 ? 1 : 0); } static int ns8250_getc(struct uart_bas *bas, struct mtx *hwmtx) { int c; uart_lock(hwmtx); while ((uart_getreg(bas, REG_LSR) & LSR_RXRDY) == 0) { uart_unlock(hwmtx); DELAY(4); uart_lock(hwmtx); } c = uart_getreg(bas, REG_DATA); uart_unlock(hwmtx); return (c); } static kobj_method_t ns8250_methods[] = { KOBJMETHOD(uart_attach, ns8250_bus_attach), KOBJMETHOD(uart_detach, ns8250_bus_detach), KOBJMETHOD(uart_flush, ns8250_bus_flush), KOBJMETHOD(uart_getsig, ns8250_bus_getsig), KOBJMETHOD(uart_ioctl, ns8250_bus_ioctl), KOBJMETHOD(uart_ipend, ns8250_bus_ipend), KOBJMETHOD(uart_param, ns8250_bus_param), KOBJMETHOD(uart_probe, ns8250_bus_probe), KOBJMETHOD(uart_receive, ns8250_bus_receive), KOBJMETHOD(uart_setsig, ns8250_bus_setsig), KOBJMETHOD(uart_transmit, ns8250_bus_transmit), KOBJMETHOD(uart_grab, ns8250_bus_grab), KOBJMETHOD(uart_ungrab, ns8250_bus_ungrab), { 0, 0 } }; struct uart_class uart_ns8250_class = { "ns8250", ns8250_methods, sizeof(struct ns8250_softc), .uc_ops = &uart_ns8250_ops, .uc_range = 8, .uc_rclk = DEFAULT_RCLK, .uc_rshift = 0 }; #ifdef FDT static struct ofw_compat_data compat_data[] = { {"ns16550", (uintptr_t)&uart_ns8250_class}, {NULL, (uintptr_t)NULL}, }; UART_FDT_CLASS_AND_DEVICE(compat_data); #endif #define SIGCHG(c, i, s, d) \ if (c) { \ i |= (i & s) ? s : s | d; \ } else { \ i = (i & s) ? (i & ~s) | d : i; \ } int ns8250_bus_attach(struct uart_softc *sc) { struct ns8250_softc *ns8250 = (struct ns8250_softc*)sc; struct uart_bas *bas; unsigned int ivar; #ifdef FDT phandle_t node; pcell_t cell; #endif ns8250->busy_detect = 0; #ifdef FDT /* * Check whether uart requires to read USR reg when IIR_BUSY and * has broken txfifo. */ node = ofw_bus_get_node(sc->sc_dev); - if ((OF_getprop(node, "busy-detect", &cell, sizeof(cell))) > 0) - ns8250->busy_detect = 1; - if ((OF_getprop(node, "broken-txfifo", &cell, sizeof(cell))) > 0) - broken_txfifo = 1; + if ((OF_getencprop(node, "busy-detect", &cell, sizeof(cell))) > 0) + ns8250->busy_detect = cell ? 1 : 0; + if ((OF_getencprop(node, "broken-txfifo", &cell, sizeof(cell))) > 0) + broken_txfifo = cell ? 1 : 0; #endif bas = &sc->sc_bas; ns8250->mcr = uart_getreg(bas, REG_MCR); ns8250->fcr = FCR_ENABLE; if (!resource_int_value("uart", device_get_unit(sc->sc_dev), "flags", &ivar)) { if (UART_FLAGS_FCR_RX_LOW(ivar)) ns8250->fcr |= FCR_RX_LOW; else if (UART_FLAGS_FCR_RX_MEDL(ivar)) ns8250->fcr |= FCR_RX_MEDL; else if (UART_FLAGS_FCR_RX_HIGH(ivar)) ns8250->fcr |= FCR_RX_HIGH; else ns8250->fcr |= FCR_RX_MEDH; } else ns8250->fcr |= FCR_RX_MEDH; /* Get IER mask */ ivar = 0xf0; resource_int_value("uart", device_get_unit(sc->sc_dev), "ier_mask", &ivar); ns8250->ier_mask = (uint8_t)(ivar & 0xff); /* Get IER RX interrupt bits */ ivar = IER_EMSC | IER_ERLS | IER_ERXRDY; resource_int_value("uart", device_get_unit(sc->sc_dev), "ier_rxbits", &ivar); ns8250->ier_rxbits = (uint8_t)(ivar & 0xff); uart_setreg(bas, REG_FCR, ns8250->fcr); uart_barrier(bas); ns8250_bus_flush(sc, UART_FLUSH_RECEIVER|UART_FLUSH_TRANSMITTER); if (ns8250->mcr & MCR_DTR) sc->sc_hwsig |= SER_DTR; if (ns8250->mcr & MCR_RTS) sc->sc_hwsig |= SER_RTS; ns8250_bus_getsig(sc); ns8250_clrint(bas); ns8250->ier = uart_getreg(bas, REG_IER) & ns8250->ier_mask; ns8250->ier |= ns8250->ier_rxbits; uart_setreg(bas, REG_IER, ns8250->ier); uart_barrier(bas); /* * Timing of the H/W access was changed with r253161 of uart_core.c * It has been observed that an ITE IT8513E would signal a break * condition with pretty much every character it received, unless * it had enough time to settle between ns8250_bus_attach() and * ns8250_bus_ipend() -- which it accidentally had before r253161. * It's not understood why the UART chip behaves this way and it * could very well be that the DELAY make the H/W work in the same * accidental manner as before. More analysis is warranted, but * at least now we fixed a known regression. */ DELAY(200); return (0); } int ns8250_bus_detach(struct uart_softc *sc) { struct ns8250_softc *ns8250; struct uart_bas *bas; u_char ier; ns8250 = (struct ns8250_softc *)sc; bas = &sc->sc_bas; ier = uart_getreg(bas, REG_IER) & ns8250->ier_mask; uart_setreg(bas, REG_IER, ier); uart_barrier(bas); ns8250_clrint(bas); return (0); } int ns8250_bus_flush(struct uart_softc *sc, int what) { struct ns8250_softc *ns8250 = (struct ns8250_softc*)sc; struct uart_bas *bas; int error; bas = &sc->sc_bas; uart_lock(sc->sc_hwmtx); if (sc->sc_rxfifosz > 1) { ns8250_flush(bas, what); uart_setreg(bas, REG_FCR, ns8250->fcr); uart_barrier(bas); error = 0; } else error = ns8250_drain(bas, what); uart_unlock(sc->sc_hwmtx); return (error); } int ns8250_bus_getsig(struct uart_softc *sc) { uint32_t new, old, sig; uint8_t msr; do { old = sc->sc_hwsig; sig = old; uart_lock(sc->sc_hwmtx); msr = uart_getreg(&sc->sc_bas, REG_MSR); uart_unlock(sc->sc_hwmtx); SIGCHG(msr & MSR_DSR, sig, SER_DSR, SER_DDSR); SIGCHG(msr & MSR_CTS, sig, SER_CTS, SER_DCTS); SIGCHG(msr & MSR_DCD, sig, SER_DCD, SER_DDCD); SIGCHG(msr & MSR_RI, sig, SER_RI, SER_DRI); new = sig & ~SER_MASK_DELTA; } while (!atomic_cmpset_32(&sc->sc_hwsig, old, new)); return (sig); } int ns8250_bus_ioctl(struct uart_softc *sc, int request, intptr_t data) { struct uart_bas *bas; int baudrate, divisor, error; uint8_t efr, lcr; bas = &sc->sc_bas; error = 0; uart_lock(sc->sc_hwmtx); switch (request) { case UART_IOCTL_BREAK: lcr = uart_getreg(bas, REG_LCR); if (data) lcr |= LCR_SBREAK; else lcr &= ~LCR_SBREAK; uart_setreg(bas, REG_LCR, lcr); uart_barrier(bas); break; case UART_IOCTL_IFLOW: lcr = uart_getreg(bas, REG_LCR); uart_barrier(bas); uart_setreg(bas, REG_LCR, 0xbf); uart_barrier(bas); efr = uart_getreg(bas, REG_EFR); if (data) efr |= EFR_RTS; else efr &= ~EFR_RTS; uart_setreg(bas, REG_EFR, efr); uart_barrier(bas); uart_setreg(bas, REG_LCR, lcr); uart_barrier(bas); break; case UART_IOCTL_OFLOW: lcr = uart_getreg(bas, REG_LCR); uart_barrier(bas); uart_setreg(bas, REG_LCR, 0xbf); uart_barrier(bas); efr = uart_getreg(bas, REG_EFR); if (data) efr |= EFR_CTS; else efr &= ~EFR_CTS; uart_setreg(bas, REG_EFR, efr); uart_barrier(bas); uart_setreg(bas, REG_LCR, lcr); uart_barrier(bas); break; case UART_IOCTL_BAUD: lcr = uart_getreg(bas, REG_LCR); uart_setreg(bas, REG_LCR, lcr | LCR_DLAB); uart_barrier(bas); divisor = uart_getreg(bas, REG_DLL) | (uart_getreg(bas, REG_DLH) << 8); uart_barrier(bas); uart_setreg(bas, REG_LCR, lcr); uart_barrier(bas); baudrate = (divisor > 0) ? bas->rclk / divisor / 16 : 0; if (baudrate > 0) *(int*)data = baudrate; else error = ENXIO; break; default: error = EINVAL; break; } uart_unlock(sc->sc_hwmtx); return (error); } int ns8250_bus_ipend(struct uart_softc *sc) { struct uart_bas *bas; struct ns8250_softc *ns8250; int ipend; uint8_t iir, lsr; ns8250 = (struct ns8250_softc *)sc; bas = &sc->sc_bas; uart_lock(sc->sc_hwmtx); iir = uart_getreg(bas, REG_IIR); if (ns8250->busy_detect && (iir & IIR_BUSY) == IIR_BUSY) { (void)uart_getreg(bas, DW_REG_USR); uart_unlock(sc->sc_hwmtx); return (0); } if (iir & IIR_NOPEND) { uart_unlock(sc->sc_hwmtx); return (0); } ipend = 0; if (iir & IIR_RXRDY) { lsr = uart_getreg(bas, REG_LSR); if (lsr & LSR_OE) ipend |= SER_INT_OVERRUN; if (lsr & LSR_BI) ipend |= SER_INT_BREAK; if (lsr & LSR_RXRDY) ipend |= SER_INT_RXREADY; } else { if (iir & IIR_TXRDY) { ipend |= SER_INT_TXIDLE; uart_setreg(bas, REG_IER, ns8250->ier); } else ipend |= SER_INT_SIGCHG; } if (ipend == 0) ns8250_clrint(bas); uart_unlock(sc->sc_hwmtx); return (ipend); } int ns8250_bus_param(struct uart_softc *sc, int baudrate, int databits, int stopbits, int parity) { struct ns8250_softc *ns8250; struct uart_bas *bas; int error, limit; ns8250 = (struct ns8250_softc*)sc; bas = &sc->sc_bas; uart_lock(sc->sc_hwmtx); /* * When using DW UART with BUSY detection it is necessary to wait * until all serial transfers are finished before manipulating the * line control. LCR will not be affected when UART is busy. */ if (ns8250->busy_detect != 0) { /* * Pick an arbitrary high limit to avoid getting stuck in * an infinite loop in case when the hardware is broken. */ limit = 10 * 1024; while (((uart_getreg(bas, DW_REG_USR) & USR_BUSY) != 0) && --limit) DELAY(4); if (limit <= 0) { /* UART appears to be stuck */ uart_unlock(sc->sc_hwmtx); return (EIO); } } error = ns8250_param(bas, baudrate, databits, stopbits, parity); uart_unlock(sc->sc_hwmtx); return (error); } int ns8250_bus_probe(struct uart_softc *sc) { struct ns8250_softc *ns8250; struct uart_bas *bas; int count, delay, error, limit; uint8_t lsr, mcr, ier; ns8250 = (struct ns8250_softc *)sc; bas = &sc->sc_bas; error = ns8250_probe(bas); if (error) return (error); mcr = MCR_IE; if (sc->sc_sysdev == NULL) { /* By using ns8250_init() we also set DTR and RTS. */ ns8250_init(bas, 115200, 8, 1, UART_PARITY_NONE); } else mcr |= MCR_DTR | MCR_RTS; error = ns8250_drain(bas, UART_DRAIN_TRANSMITTER); if (error) return (error); /* * Set loopback mode. This avoids having garbage on the wire and * also allows us send and receive data. We set DTR and RTS to * avoid the possibility that automatic flow-control prevents * any data from being sent. */ uart_setreg(bas, REG_MCR, MCR_LOOPBACK | MCR_IE | MCR_DTR | MCR_RTS); uart_barrier(bas); /* * Enable FIFOs. And check that the UART has them. If not, we're * done. Since this is the first time we enable the FIFOs, we reset * them. */ uart_setreg(bas, REG_FCR, FCR_ENABLE); uart_barrier(bas); if (!(uart_getreg(bas, REG_IIR) & IIR_FIFO_MASK)) { /* * NS16450 or INS8250. We don't bother to differentiate * between them. They're too old to be interesting. */ uart_setreg(bas, REG_MCR, mcr); uart_barrier(bas); sc->sc_rxfifosz = sc->sc_txfifosz = 1; device_set_desc(sc->sc_dev, "8250 or 16450 or compatible"); return (0); } uart_setreg(bas, REG_FCR, FCR_ENABLE | FCR_XMT_RST | FCR_RCV_RST); uart_barrier(bas); count = 0; delay = ns8250_delay(bas); /* We have FIFOs. Drain the transmitter and receiver. */ error = ns8250_drain(bas, UART_DRAIN_RECEIVER|UART_DRAIN_TRANSMITTER); if (error) { uart_setreg(bas, REG_MCR, mcr); uart_setreg(bas, REG_FCR, 0); uart_barrier(bas); goto describe; } /* * We should have a sufficiently clean "pipe" to determine the * size of the FIFOs. We send as much characters as is reasonable * and wait for the overflow bit in the LSR register to be * asserted, counting the characters as we send them. Based on * that count we know the FIFO size. */ do { uart_setreg(bas, REG_DATA, 0); uart_barrier(bas); count++; limit = 30; lsr = 0; /* * LSR bits are cleared upon read, so we must accumulate * them to be able to test LSR_OE below. */ while (((lsr |= uart_getreg(bas, REG_LSR)) & LSR_TEMT) == 0 && --limit) DELAY(delay); if (limit == 0) { ier = uart_getreg(bas, REG_IER) & ns8250->ier_mask; uart_setreg(bas, REG_IER, ier); uart_setreg(bas, REG_MCR, mcr); uart_setreg(bas, REG_FCR, 0); uart_barrier(bas); count = 0; goto describe; } } while ((lsr & LSR_OE) == 0 && count < 130); count--; uart_setreg(bas, REG_MCR, mcr); /* Reset FIFOs. */ ns8250_flush(bas, UART_FLUSH_RECEIVER|UART_FLUSH_TRANSMITTER); describe: if (count >= 14 && count <= 16) { sc->sc_rxfifosz = 16; device_set_desc(sc->sc_dev, "16550 or compatible"); } else if (count >= 28 && count <= 32) { sc->sc_rxfifosz = 32; device_set_desc(sc->sc_dev, "16650 or compatible"); } else if (count >= 56 && count <= 64) { sc->sc_rxfifosz = 64; device_set_desc(sc->sc_dev, "16750 or compatible"); } else if (count >= 112 && count <= 128) { sc->sc_rxfifosz = 128; device_set_desc(sc->sc_dev, "16950 or compatible"); } else { sc->sc_rxfifosz = 16; device_set_desc(sc->sc_dev, "Non-standard ns8250 class UART with FIFOs"); } /* * Force the Tx FIFO size to 16 bytes for now. We don't program the * Tx trigger. Also, we assume that all data has been sent when the * interrupt happens. */ sc->sc_txfifosz = 16; #if 0 /* * XXX there are some issues related to hardware flow control and * it's likely that uart(4) is the cause. This basicly needs more * investigation, but we avoid using for hardware flow control * until then. */ /* 16650s or higher have automatic flow control. */ if (sc->sc_rxfifosz > 16) { sc->sc_hwiflow = 1; sc->sc_hwoflow = 1; } #endif return (0); } int ns8250_bus_receive(struct uart_softc *sc) { struct uart_bas *bas; int xc; uint8_t lsr; bas = &sc->sc_bas; uart_lock(sc->sc_hwmtx); lsr = uart_getreg(bas, REG_LSR); while (lsr & LSR_RXRDY) { if (uart_rx_full(sc)) { sc->sc_rxbuf[sc->sc_rxput] = UART_STAT_OVERRUN; break; } xc = uart_getreg(bas, REG_DATA); if (lsr & LSR_FE) xc |= UART_STAT_FRAMERR; if (lsr & LSR_PE) xc |= UART_STAT_PARERR; uart_rx_put(sc, xc); lsr = uart_getreg(bas, REG_LSR); } /* Discard everything left in the Rx FIFO. */ while (lsr & LSR_RXRDY) { (void)uart_getreg(bas, REG_DATA); uart_barrier(bas); lsr = uart_getreg(bas, REG_LSR); } uart_unlock(sc->sc_hwmtx); return (0); } int ns8250_bus_setsig(struct uart_softc *sc, int sig) { struct ns8250_softc *ns8250 = (struct ns8250_softc*)sc; struct uart_bas *bas; uint32_t new, old; bas = &sc->sc_bas; do { old = sc->sc_hwsig; new = old; if (sig & SER_DDTR) { SIGCHG(sig & SER_DTR, new, SER_DTR, SER_DDTR); } if (sig & SER_DRTS) { SIGCHG(sig & SER_RTS, new, SER_RTS, SER_DRTS); } } while (!atomic_cmpset_32(&sc->sc_hwsig, old, new)); uart_lock(sc->sc_hwmtx); ns8250->mcr &= ~(MCR_DTR|MCR_RTS); if (new & SER_DTR) ns8250->mcr |= MCR_DTR; if (new & SER_RTS) ns8250->mcr |= MCR_RTS; uart_setreg(bas, REG_MCR, ns8250->mcr); uart_barrier(bas); uart_unlock(sc->sc_hwmtx); return (0); } int ns8250_bus_transmit(struct uart_softc *sc) { struct ns8250_softc *ns8250 = (struct ns8250_softc*)sc; struct uart_bas *bas; int i; bas = &sc->sc_bas; uart_lock(sc->sc_hwmtx); while ((uart_getreg(bas, REG_LSR) & LSR_THRE) == 0) ; uart_setreg(bas, REG_IER, ns8250->ier | IER_ETXRDY); uart_barrier(bas); for (i = 0; i < sc->sc_txdatasz; i++) { uart_setreg(bas, REG_DATA, sc->sc_txbuf[i]); uart_barrier(bas); } if (broken_txfifo) ns8250_drain(bas, UART_DRAIN_TRANSMITTER); else sc->sc_txbusy = 1; uart_unlock(sc->sc_hwmtx); if (broken_txfifo) uart_sched_softih(sc, SER_INT_TXIDLE); return (0); } void ns8250_bus_grab(struct uart_softc *sc) { struct uart_bas *bas = &sc->sc_bas; struct ns8250_softc *ns8250 = (struct ns8250_softc*)sc; u_char ier; /* * turn off all interrupts to enter polling mode. Leave the * saved mask alone. We'll restore whatever it was in ungrab. * All pending interupt signals are reset when IER is set to 0. */ uart_lock(sc->sc_hwmtx); ier = uart_getreg(bas, REG_IER); uart_setreg(bas, REG_IER, ier & ns8250->ier_mask); uart_barrier(bas); uart_unlock(sc->sc_hwmtx); } void ns8250_bus_ungrab(struct uart_softc *sc) { struct ns8250_softc *ns8250 = (struct ns8250_softc*)sc; struct uart_bas *bas = &sc->sc_bas; /* * Restore previous interrupt mask */ uart_lock(sc->sc_hwmtx); uart_setreg(bas, REG_IER, ns8250->ier); uart_barrier(bas); uart_unlock(sc->sc_hwmtx); } Index: projects/clang-trunk/sys/libkern/arm/memcpy.S =================================================================== --- projects/clang-trunk/sys/libkern/arm/memcpy.S (revision 283870) +++ projects/clang-trunk/sys/libkern/arm/memcpy.S (revision 283871) @@ -1,38 +1,42 @@ /* * Copyright (C) 2013 Andrew Turner * All rights reserved. * * Redistribution and use in source and binary forms, with or without * modification, are permitted provided that the following conditions * are met: * 1. Redistributions of source code must retain the above copyright * notice, this list of conditions and the following disclaimer. * 2. Redistributions in binary form must reproduce the above copyright * notice, this list of conditions and the following disclaimer in the * documentation and/or other materials provided with the distribution. * * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE * ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF * SUCH DAMAGE. * */ #include __FBSDID("$FreeBSD$"); #ifdef __ARM_EABI__ ENTRY_NP(__aeabi_memcpy) +EENTRY_NP(__aeabi_memcpy4) +EENTRY_NP(__aeabi_memcpy8) b memcpy +EEND(__aeabi_memcpy8) +EEND(__aeabi_memcpy4) END(__aeabi_memcpy) #endif Index: projects/clang-trunk/sys/ufs/ffs/ffs_softdep.c =================================================================== --- projects/clang-trunk/sys/ufs/ffs/ffs_softdep.c (revision 283870) +++ projects/clang-trunk/sys/ufs/ffs/ffs_softdep.c (revision 283871) @@ -1,14245 +1,14243 @@ /*- * Copyright 1998, 2000 Marshall Kirk McKusick. * Copyright 2009, 2010 Jeffrey W. Roberson * All rights reserved. * * The soft updates code is derived from the appendix of a University * of Michigan technical report (Gregory R. Ganger and Yale N. Patt, * "Soft Updates: A Solution to the Metadata Update Problem in File * Systems", CSE-TR-254-95, August 1995). * * Further information about soft updates can be obtained from: * * Marshall Kirk McKusick http://www.mckusick.com/softdep/ * 1614 Oxford Street mckusick@mckusick.com * Berkeley, CA 94709-1608 +1-510-843-9542 * USA * * Redistribution and use in source and binary forms, with or without * modification, are permitted provided that the following conditions * are met: * * 1. Redistributions of source code must retain the above copyright * notice, this list of conditions and the following disclaimer. * 2. Redistributions in binary form must reproduce the above copyright * notice, this list of conditions and the following disclaimer in the * documentation and/or other materials provided with the distribution. * * THIS SOFTWARE IS PROVIDED BY THE AUTHORS ``AS IS'' AND ANY EXPRESS OR * IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES * OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. * IN NO EVENT SHALL THE AUTHORS BE LIABLE FOR ANY DIRECT, INDIRECT, * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, * BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS * OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND * ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR * TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE * USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. * * from: @(#)ffs_softdep.c 9.59 (McKusick) 6/21/00 */ #include __FBSDID("$FreeBSD$"); #include "opt_ffs.h" #include "opt_quota.h" #include "opt_ddb.h" /* * For now we want the safety net that the DEBUG flag provides. */ #ifndef DEBUG #define DEBUG #endif #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #define KTR_SUJ 0 /* Define to KTR_SPARE. */ #ifndef SOFTUPDATES int softdep_flushfiles(oldmnt, flags, td) struct mount *oldmnt; int flags; struct thread *td; { panic("softdep_flushfiles called"); } int softdep_mount(devvp, mp, fs, cred) struct vnode *devvp; struct mount *mp; struct fs *fs; struct ucred *cred; { return (0); } void softdep_initialize() { return; } void softdep_uninitialize() { return; } void softdep_unmount(mp) struct mount *mp; { panic("softdep_unmount called"); } void softdep_setup_sbupdate(ump, fs, bp) struct ufsmount *ump; struct fs *fs; struct buf *bp; { panic("softdep_setup_sbupdate called"); } void softdep_setup_inomapdep(bp, ip, newinum, mode) struct buf *bp; struct inode *ip; ino_t newinum; int mode; { panic("softdep_setup_inomapdep called"); } void softdep_setup_blkmapdep(bp, mp, newblkno, frags, oldfrags) struct buf *bp; struct mount *mp; ufs2_daddr_t newblkno; int frags; int oldfrags; { panic("softdep_setup_blkmapdep called"); } void softdep_setup_allocdirect(ip, lbn, newblkno, oldblkno, newsize, oldsize, bp) struct inode *ip; ufs_lbn_t lbn; ufs2_daddr_t newblkno; ufs2_daddr_t oldblkno; long newsize; long oldsize; struct buf *bp; { panic("softdep_setup_allocdirect called"); } void softdep_setup_allocext(ip, lbn, newblkno, oldblkno, newsize, oldsize, bp) struct inode *ip; ufs_lbn_t lbn; ufs2_daddr_t newblkno; ufs2_daddr_t oldblkno; long newsize; long oldsize; struct buf *bp; { panic("softdep_setup_allocext called"); } void softdep_setup_allocindir_page(ip, lbn, bp, ptrno, newblkno, oldblkno, nbp) struct inode *ip; ufs_lbn_t lbn; struct buf *bp; int ptrno; ufs2_daddr_t newblkno; ufs2_daddr_t oldblkno; struct buf *nbp; { panic("softdep_setup_allocindir_page called"); } void softdep_setup_allocindir_meta(nbp, ip, bp, ptrno, newblkno) struct buf *nbp; struct inode *ip; struct buf *bp; int ptrno; ufs2_daddr_t newblkno; { panic("softdep_setup_allocindir_meta called"); } void softdep_journal_freeblocks(ip, cred, length, flags) struct inode *ip; struct ucred *cred; off_t length; int flags; { panic("softdep_journal_freeblocks called"); } void softdep_journal_fsync(ip) struct inode *ip; { panic("softdep_journal_fsync called"); } void softdep_setup_freeblocks(ip, length, flags) struct inode *ip; off_t length; int flags; { panic("softdep_setup_freeblocks called"); } void softdep_freefile(pvp, ino, mode) struct vnode *pvp; ino_t ino; int mode; { panic("softdep_freefile called"); } int softdep_setup_directory_add(bp, dp, diroffset, newinum, newdirbp, isnewblk) struct buf *bp; struct inode *dp; off_t diroffset; ino_t newinum; struct buf *newdirbp; int isnewblk; { panic("softdep_setup_directory_add called"); } void softdep_change_directoryentry_offset(bp, dp, base, oldloc, newloc, entrysize) struct buf *bp; struct inode *dp; caddr_t base; caddr_t oldloc; caddr_t newloc; int entrysize; { panic("softdep_change_directoryentry_offset called"); } void softdep_setup_remove(bp, dp, ip, isrmdir) struct buf *bp; struct inode *dp; struct inode *ip; int isrmdir; { panic("softdep_setup_remove called"); } void softdep_setup_directory_change(bp, dp, ip, newinum, isrmdir) struct buf *bp; struct inode *dp; struct inode *ip; ino_t newinum; int isrmdir; { panic("softdep_setup_directory_change called"); } void softdep_setup_blkfree(mp, bp, blkno, frags, wkhd) struct mount *mp; struct buf *bp; ufs2_daddr_t blkno; int frags; struct workhead *wkhd; { panic("%s called", __FUNCTION__); } void softdep_setup_inofree(mp, bp, ino, wkhd) struct mount *mp; struct buf *bp; ino_t ino; struct workhead *wkhd; { panic("%s called", __FUNCTION__); } void softdep_setup_unlink(dp, ip) struct inode *dp; struct inode *ip; { panic("%s called", __FUNCTION__); } void softdep_setup_link(dp, ip) struct inode *dp; struct inode *ip; { panic("%s called", __FUNCTION__); } void softdep_revert_link(dp, ip) struct inode *dp; struct inode *ip; { panic("%s called", __FUNCTION__); } void softdep_setup_rmdir(dp, ip) struct inode *dp; struct inode *ip; { panic("%s called", __FUNCTION__); } void softdep_revert_rmdir(dp, ip) struct inode *dp; struct inode *ip; { panic("%s called", __FUNCTION__); } void softdep_setup_create(dp, ip) struct inode *dp; struct inode *ip; { panic("%s called", __FUNCTION__); } void softdep_revert_create(dp, ip) struct inode *dp; struct inode *ip; { panic("%s called", __FUNCTION__); } void softdep_setup_mkdir(dp, ip) struct inode *dp; struct inode *ip; { panic("%s called", __FUNCTION__); } void softdep_revert_mkdir(dp, ip) struct inode *dp; struct inode *ip; { panic("%s called", __FUNCTION__); } void softdep_setup_dotdot_link(dp, ip) struct inode *dp; struct inode *ip; { panic("%s called", __FUNCTION__); } int softdep_prealloc(vp, waitok) struct vnode *vp; int waitok; { panic("%s called", __FUNCTION__); } int softdep_journal_lookup(mp, vpp) struct mount *mp; struct vnode **vpp; { return (ENOENT); } void softdep_change_linkcnt(ip) struct inode *ip; { panic("softdep_change_linkcnt called"); } void softdep_load_inodeblock(ip) struct inode *ip; { panic("softdep_load_inodeblock called"); } void softdep_update_inodeblock(ip, bp, waitfor) struct inode *ip; struct buf *bp; int waitfor; { panic("softdep_update_inodeblock called"); } int softdep_fsync(vp) struct vnode *vp; /* the "in_core" copy of the inode */ { return (0); } void softdep_fsync_mountdev(vp) struct vnode *vp; { return; } int softdep_flushworklist(oldmnt, countp, td) struct mount *oldmnt; int *countp; struct thread *td; { *countp = 0; return (0); } int softdep_sync_metadata(struct vnode *vp) { panic("softdep_sync_metadata called"); } int softdep_sync_buf(struct vnode *vp, struct buf *bp, int waitfor) { panic("softdep_sync_buf called"); } int softdep_slowdown(vp) struct vnode *vp; { panic("softdep_slowdown called"); } int softdep_request_cleanup(fs, vp, cred, resource) struct fs *fs; struct vnode *vp; struct ucred *cred; int resource; { return (0); } int softdep_check_suspend(struct mount *mp, struct vnode *devvp, int softdep_depcnt, int softdep_accdepcnt, int secondary_writes, int secondary_accwrites) { struct bufobj *bo; int error; (void) softdep_depcnt, (void) softdep_accdepcnt; bo = &devvp->v_bufobj; ASSERT_BO_WLOCKED(bo); MNT_ILOCK(mp); while (mp->mnt_secondary_writes != 0) { BO_UNLOCK(bo); msleep(&mp->mnt_secondary_writes, MNT_MTX(mp), (PUSER - 1) | PDROP, "secwr", 0); BO_LOCK(bo); MNT_ILOCK(mp); } /* * Reasons for needing more work before suspend: * - Dirty buffers on devvp. * - Secondary writes occurred after start of vnode sync loop */ error = 0; if (bo->bo_numoutput > 0 || bo->bo_dirty.bv_cnt > 0 || secondary_writes != 0 || mp->mnt_secondary_writes != 0 || secondary_accwrites != mp->mnt_secondary_accwrites) error = EAGAIN; BO_UNLOCK(bo); return (error); } void softdep_get_depcounts(struct mount *mp, int *softdepactivep, int *softdepactiveaccp) { (void) mp; *softdepactivep = 0; *softdepactiveaccp = 0; } void softdep_buf_append(bp, wkhd) struct buf *bp; struct workhead *wkhd; { panic("softdep_buf_appendwork called"); } void softdep_inode_append(ip, cred, wkhd) struct inode *ip; struct ucred *cred; struct workhead *wkhd; { panic("softdep_inode_appendwork called"); } void softdep_freework(wkhd) struct workhead *wkhd; { panic("softdep_freework called"); } #else FEATURE(softupdates, "FFS soft-updates support"); static SYSCTL_NODE(_debug, OID_AUTO, softdep, CTLFLAG_RW, 0, "soft updates stats"); static SYSCTL_NODE(_debug_softdep, OID_AUTO, total, CTLFLAG_RW, 0, "total dependencies allocated"); static SYSCTL_NODE(_debug_softdep, OID_AUTO, highuse, CTLFLAG_RW, 0, "high use dependencies allocated"); static SYSCTL_NODE(_debug_softdep, OID_AUTO, current, CTLFLAG_RW, 0, "current dependencies allocated"); static SYSCTL_NODE(_debug_softdep, OID_AUTO, write, CTLFLAG_RW, 0, "current dependencies written"); unsigned long dep_current[D_LAST + 1]; unsigned long dep_highuse[D_LAST + 1]; unsigned long dep_total[D_LAST + 1]; unsigned long dep_write[D_LAST + 1]; #define SOFTDEP_TYPE(type, str, long) \ static MALLOC_DEFINE(M_ ## type, #str, long); \ SYSCTL_ULONG(_debug_softdep_total, OID_AUTO, str, CTLFLAG_RD, \ &dep_total[D_ ## type], 0, ""); \ SYSCTL_ULONG(_debug_softdep_current, OID_AUTO, str, CTLFLAG_RD, \ &dep_current[D_ ## type], 0, ""); \ SYSCTL_ULONG(_debug_softdep_highuse, OID_AUTO, str, CTLFLAG_RD, \ &dep_highuse[D_ ## type], 0, ""); \ SYSCTL_ULONG(_debug_softdep_write, OID_AUTO, str, CTLFLAG_RD, \ &dep_write[D_ ## type], 0, ""); SOFTDEP_TYPE(PAGEDEP, pagedep, "File page dependencies"); SOFTDEP_TYPE(INODEDEP, inodedep, "Inode dependencies"); SOFTDEP_TYPE(BMSAFEMAP, bmsafemap, "Block or frag allocated from cyl group map"); SOFTDEP_TYPE(NEWBLK, newblk, "New block or frag allocation dependency"); SOFTDEP_TYPE(ALLOCDIRECT, allocdirect, "Block or frag dependency for an inode"); SOFTDEP_TYPE(INDIRDEP, indirdep, "Indirect block dependencies"); SOFTDEP_TYPE(ALLOCINDIR, allocindir, "Block dependency for an indirect block"); SOFTDEP_TYPE(FREEFRAG, freefrag, "Previously used frag for an inode"); SOFTDEP_TYPE(FREEBLKS, freeblks, "Blocks freed from an inode"); SOFTDEP_TYPE(FREEFILE, freefile, "Inode deallocated"); SOFTDEP_TYPE(DIRADD, diradd, "New directory entry"); SOFTDEP_TYPE(MKDIR, mkdir, "New directory"); SOFTDEP_TYPE(DIRREM, dirrem, "Directory entry deleted"); SOFTDEP_TYPE(NEWDIRBLK, newdirblk, "Unclaimed new directory block"); SOFTDEP_TYPE(FREEWORK, freework, "free an inode block"); SOFTDEP_TYPE(FREEDEP, freedep, "track a block free"); SOFTDEP_TYPE(JADDREF, jaddref, "Journal inode ref add"); SOFTDEP_TYPE(JREMREF, jremref, "Journal inode ref remove"); SOFTDEP_TYPE(JMVREF, jmvref, "Journal inode ref move"); SOFTDEP_TYPE(JNEWBLK, jnewblk, "Journal new block"); SOFTDEP_TYPE(JFREEBLK, jfreeblk, "Journal free block"); SOFTDEP_TYPE(JFREEFRAG, jfreefrag, "Journal free frag"); SOFTDEP_TYPE(JSEG, jseg, "Journal segment"); SOFTDEP_TYPE(JSEGDEP, jsegdep, "Journal segment complete"); SOFTDEP_TYPE(SBDEP, sbdep, "Superblock write dependency"); SOFTDEP_TYPE(JTRUNC, jtrunc, "Journal inode truncation"); SOFTDEP_TYPE(JFSYNC, jfsync, "Journal fsync complete"); static MALLOC_DEFINE(M_SENTINEL, "sentinel", "Worklist sentinel"); static MALLOC_DEFINE(M_SAVEDINO, "savedino", "Saved inodes"); static MALLOC_DEFINE(M_JBLOCKS, "jblocks", "Journal block locations"); static MALLOC_DEFINE(M_MOUNTDATA, "softdep", "Softdep per-mount data"); #define M_SOFTDEP_FLAGS (M_WAITOK) /* * translate from workitem type to memory type * MUST match the defines above, such that memtype[D_XXX] == M_XXX */ static struct malloc_type *memtype[] = { M_PAGEDEP, M_INODEDEP, M_BMSAFEMAP, M_NEWBLK, M_ALLOCDIRECT, M_INDIRDEP, M_ALLOCINDIR, M_FREEFRAG, M_FREEBLKS, M_FREEFILE, M_DIRADD, M_MKDIR, M_DIRREM, M_NEWDIRBLK, M_FREEWORK, M_FREEDEP, M_JADDREF, M_JREMREF, M_JMVREF, M_JNEWBLK, M_JFREEBLK, M_JFREEFRAG, M_JSEG, M_JSEGDEP, M_SBDEP, M_JTRUNC, M_JFSYNC, M_SENTINEL }; #define DtoM(type) (memtype[type]) /* * Names of malloc types. */ #define TYPENAME(type) \ ((unsigned)(type) <= D_LAST ? memtype[type]->ks_shortdesc : "???") /* * End system adaptation definitions. */ #define DOTDOT_OFFSET offsetof(struct dirtemplate, dotdot_ino) #define DOT_OFFSET offsetof(struct dirtemplate, dot_ino) /* * Internal function prototypes. */ static void check_clear_deps(struct mount *); static void softdep_error(char *, int); static int softdep_process_worklist(struct mount *, int); static int softdep_waitidle(struct mount *, int); static void drain_output(struct vnode *); static struct buf *getdirtybuf(struct buf *, struct rwlock *, int); static int check_inodedep_free(struct inodedep *); static void clear_remove(struct mount *); static void clear_inodedeps(struct mount *); static void unlinked_inodedep(struct mount *, struct inodedep *); static void clear_unlinked_inodedep(struct inodedep *); static struct inodedep *first_unlinked_inodedep(struct ufsmount *); static int flush_pagedep_deps(struct vnode *, struct mount *, struct diraddhd *); static int free_pagedep(struct pagedep *); static int flush_newblk_dep(struct vnode *, struct mount *, ufs_lbn_t); static int flush_inodedep_deps(struct vnode *, struct mount *, ino_t); static int flush_deplist(struct allocdirectlst *, int, int *); static int sync_cgs(struct mount *, int); static int handle_written_filepage(struct pagedep *, struct buf *); static int handle_written_sbdep(struct sbdep *, struct buf *); static void initiate_write_sbdep(struct sbdep *); static void diradd_inode_written(struct diradd *, struct inodedep *); static int handle_written_indirdep(struct indirdep *, struct buf *, struct buf**); static int handle_written_inodeblock(struct inodedep *, struct buf *); static int jnewblk_rollforward(struct jnewblk *, struct fs *, struct cg *, uint8_t *); static int handle_written_bmsafemap(struct bmsafemap *, struct buf *); static void handle_written_jaddref(struct jaddref *); static void handle_written_jremref(struct jremref *); static void handle_written_jseg(struct jseg *, struct buf *); static void handle_written_jnewblk(struct jnewblk *); static void handle_written_jblkdep(struct jblkdep *); static void handle_written_jfreefrag(struct jfreefrag *); static void complete_jseg(struct jseg *); static void complete_jsegs(struct jseg *); static void jseg_write(struct ufsmount *ump, struct jseg *, uint8_t *); static void jaddref_write(struct jaddref *, struct jseg *, uint8_t *); static void jremref_write(struct jremref *, struct jseg *, uint8_t *); static void jmvref_write(struct jmvref *, struct jseg *, uint8_t *); static void jtrunc_write(struct jtrunc *, struct jseg *, uint8_t *); static void jfsync_write(struct jfsync *, struct jseg *, uint8_t *data); static void jnewblk_write(struct jnewblk *, struct jseg *, uint8_t *); static void jfreeblk_write(struct jfreeblk *, struct jseg *, uint8_t *); static void jfreefrag_write(struct jfreefrag *, struct jseg *, uint8_t *); static inline void inoref_write(struct inoref *, struct jseg *, struct jrefrec *); static void handle_allocdirect_partdone(struct allocdirect *, struct workhead *); static struct jnewblk *cancel_newblk(struct newblk *, struct worklist *, struct workhead *); static void indirdep_complete(struct indirdep *); static int indirblk_lookup(struct mount *, ufs2_daddr_t); static void indirblk_insert(struct freework *); static void indirblk_remove(struct freework *); static void handle_allocindir_partdone(struct allocindir *); static void initiate_write_filepage(struct pagedep *, struct buf *); static void initiate_write_indirdep(struct indirdep*, struct buf *); static void handle_written_mkdir(struct mkdir *, int); static int jnewblk_rollback(struct jnewblk *, struct fs *, struct cg *, uint8_t *); static void initiate_write_bmsafemap(struct bmsafemap *, struct buf *); static void initiate_write_inodeblock_ufs1(struct inodedep *, struct buf *); static void initiate_write_inodeblock_ufs2(struct inodedep *, struct buf *); static void handle_workitem_freefile(struct freefile *); static int handle_workitem_remove(struct dirrem *, int); static struct dirrem *newdirrem(struct buf *, struct inode *, struct inode *, int, struct dirrem **); static struct indirdep *indirdep_lookup(struct mount *, struct inode *, struct buf *); static void cancel_indirdep(struct indirdep *, struct buf *, struct freeblks *); static void free_indirdep(struct indirdep *); static void free_diradd(struct diradd *, struct workhead *); static void merge_diradd(struct inodedep *, struct diradd *); static void complete_diradd(struct diradd *); static struct diradd *diradd_lookup(struct pagedep *, int); static struct jremref *cancel_diradd_dotdot(struct inode *, struct dirrem *, struct jremref *); static struct jremref *cancel_mkdir_dotdot(struct inode *, struct dirrem *, struct jremref *); static void cancel_diradd(struct diradd *, struct dirrem *, struct jremref *, struct jremref *, struct jremref *); static void dirrem_journal(struct dirrem *, struct jremref *, struct jremref *, struct jremref *); static void cancel_allocindir(struct allocindir *, struct buf *bp, struct freeblks *, int); static int setup_trunc_indir(struct freeblks *, struct inode *, ufs_lbn_t, ufs_lbn_t, ufs2_daddr_t); static void complete_trunc_indir(struct freework *); static void trunc_indirdep(struct indirdep *, struct freeblks *, struct buf *, int); static void complete_mkdir(struct mkdir *); static void free_newdirblk(struct newdirblk *); static void free_jremref(struct jremref *); static void free_jaddref(struct jaddref *); static void free_jsegdep(struct jsegdep *); static void free_jsegs(struct jblocks *); static void rele_jseg(struct jseg *); static void free_jseg(struct jseg *, struct jblocks *); static void free_jnewblk(struct jnewblk *); static void free_jblkdep(struct jblkdep *); static void free_jfreefrag(struct jfreefrag *); static void free_freedep(struct freedep *); static void journal_jremref(struct dirrem *, struct jremref *, struct inodedep *); static void cancel_jnewblk(struct jnewblk *, struct workhead *); static int cancel_jaddref(struct jaddref *, struct inodedep *, struct workhead *); static void cancel_jfreefrag(struct jfreefrag *); static inline void setup_freedirect(struct freeblks *, struct inode *, int, int); static inline void setup_freeext(struct freeblks *, struct inode *, int, int); static inline void setup_freeindir(struct freeblks *, struct inode *, int, ufs_lbn_t, int); static inline struct freeblks *newfreeblks(struct mount *, struct inode *); static void freeblks_free(struct ufsmount *, struct freeblks *, int); static void indir_trunc(struct freework *, ufs2_daddr_t, ufs_lbn_t); static ufs2_daddr_t blkcount(struct fs *, ufs2_daddr_t, off_t); static int trunc_check_buf(struct buf *, int *, ufs_lbn_t, int, int); static void trunc_dependencies(struct inode *, struct freeblks *, ufs_lbn_t, int, int); static void trunc_pages(struct inode *, off_t, ufs2_daddr_t, int); static int cancel_pagedep(struct pagedep *, struct freeblks *, int); static int deallocate_dependencies(struct buf *, struct freeblks *, int); static void newblk_freefrag(struct newblk*); static void free_newblk(struct newblk *); static void cancel_allocdirect(struct allocdirectlst *, struct allocdirect *, struct freeblks *); static int check_inode_unwritten(struct inodedep *); static int free_inodedep(struct inodedep *); static void freework_freeblock(struct freework *); static void freework_enqueue(struct freework *); static int handle_workitem_freeblocks(struct freeblks *, int); static int handle_complete_freeblocks(struct freeblks *, int); static void handle_workitem_indirblk(struct freework *); static void handle_written_freework(struct freework *); static void merge_inode_lists(struct allocdirectlst *,struct allocdirectlst *); static struct worklist *jnewblk_merge(struct worklist *, struct worklist *, struct workhead *); static struct freefrag *setup_allocindir_phase2(struct buf *, struct inode *, struct inodedep *, struct allocindir *, ufs_lbn_t); static struct allocindir *newallocindir(struct inode *, int, ufs2_daddr_t, ufs2_daddr_t, ufs_lbn_t); static void handle_workitem_freefrag(struct freefrag *); static struct freefrag *newfreefrag(struct inode *, ufs2_daddr_t, long, ufs_lbn_t); static void allocdirect_merge(struct allocdirectlst *, struct allocdirect *, struct allocdirect *); static struct freefrag *allocindir_merge(struct allocindir *, struct allocindir *); static int bmsafemap_find(struct bmsafemap_hashhead *, int, struct bmsafemap **); static struct bmsafemap *bmsafemap_lookup(struct mount *, struct buf *, int cg, struct bmsafemap *); static int newblk_find(struct newblk_hashhead *, ufs2_daddr_t, int, struct newblk **); static int newblk_lookup(struct mount *, ufs2_daddr_t, int, struct newblk **); static int inodedep_find(struct inodedep_hashhead *, ino_t, struct inodedep **); static int inodedep_lookup(struct mount *, ino_t, int, struct inodedep **); static int pagedep_lookup(struct mount *, struct buf *bp, ino_t, ufs_lbn_t, int, struct pagedep **); static int pagedep_find(struct pagedep_hashhead *, ino_t, ufs_lbn_t, struct pagedep **); static void pause_timer(void *); static int request_cleanup(struct mount *, int); static void schedule_cleanup(struct mount *); static void softdep_ast_cleanup_proc(void); static int process_worklist_item(struct mount *, int, int); static void process_removes(struct vnode *); static void process_truncates(struct vnode *); static void jwork_move(struct workhead *, struct workhead *); static void jwork_insert(struct workhead *, struct jsegdep *); static void add_to_worklist(struct worklist *, int); static void wake_worklist(struct worklist *); static void wait_worklist(struct worklist *, char *); static void remove_from_worklist(struct worklist *); static void softdep_flush(void *); static void softdep_flushjournal(struct mount *); static int softdep_speedup(struct ufsmount *); static void worklist_speedup(struct mount *); static int journal_mount(struct mount *, struct fs *, struct ucred *); static void journal_unmount(struct ufsmount *); static int journal_space(struct ufsmount *, int); static void journal_suspend(struct ufsmount *); static int journal_unsuspend(struct ufsmount *ump); static void softdep_prelink(struct vnode *, struct vnode *); static void add_to_journal(struct worklist *); static void remove_from_journal(struct worklist *); static bool softdep_excess_inodes(struct ufsmount *); static bool softdep_excess_dirrem(struct ufsmount *); static void softdep_process_journal(struct mount *, struct worklist *, int); static struct jremref *newjremref(struct dirrem *, struct inode *, struct inode *ip, off_t, nlink_t); static struct jaddref *newjaddref(struct inode *, ino_t, off_t, int16_t, uint16_t); static inline void newinoref(struct inoref *, ino_t, ino_t, off_t, nlink_t, uint16_t); static inline struct jsegdep *inoref_jseg(struct inoref *); static struct jmvref *newjmvref(struct inode *, ino_t, off_t, off_t); static struct jfreeblk *newjfreeblk(struct freeblks *, ufs_lbn_t, ufs2_daddr_t, int); static void adjust_newfreework(struct freeblks *, int); static struct jtrunc *newjtrunc(struct freeblks *, off_t, int); static void move_newblock_dep(struct jaddref *, struct inodedep *); static void cancel_jfreeblk(struct freeblks *, ufs2_daddr_t); static struct jfreefrag *newjfreefrag(struct freefrag *, struct inode *, ufs2_daddr_t, long, ufs_lbn_t); static struct freework *newfreework(struct ufsmount *, struct freeblks *, struct freework *, ufs_lbn_t, ufs2_daddr_t, int, int, int); static int jwait(struct worklist *, int); static struct inodedep *inodedep_lookup_ip(struct inode *); static int bmsafemap_backgroundwrite(struct bmsafemap *, struct buf *); static struct freefile *handle_bufwait(struct inodedep *, struct workhead *); static void handle_jwork(struct workhead *); static struct mkdir *setup_newdir(struct diradd *, ino_t, ino_t, struct buf *, struct mkdir **); static struct jblocks *jblocks_create(void); static ufs2_daddr_t jblocks_alloc(struct jblocks *, int, int *); static void jblocks_free(struct jblocks *, struct mount *, int); static void jblocks_destroy(struct jblocks *); static void jblocks_add(struct jblocks *, ufs2_daddr_t, int); /* * Exported softdep operations. */ static void softdep_disk_io_initiation(struct buf *); static void softdep_disk_write_complete(struct buf *); static void softdep_deallocate_dependencies(struct buf *); static int softdep_count_dependencies(struct buf *bp, int); /* * Global lock over all of soft updates. */ static struct mtx lk; MTX_SYSINIT(softdep_lock, &lk, "Global Softdep Lock", MTX_DEF); #define ACQUIRE_GBLLOCK(lk) mtx_lock(lk) #define FREE_GBLLOCK(lk) mtx_unlock(lk) #define GBLLOCK_OWNED(lk) mtx_assert((lk), MA_OWNED) /* * Per-filesystem soft-updates locking. */ #define LOCK_PTR(ump) (&(ump)->um_softdep->sd_fslock) #define TRY_ACQUIRE_LOCK(ump) rw_try_wlock(&(ump)->um_softdep->sd_fslock) #define ACQUIRE_LOCK(ump) rw_wlock(&(ump)->um_softdep->sd_fslock) #define FREE_LOCK(ump) rw_wunlock(&(ump)->um_softdep->sd_fslock) #define LOCK_OWNED(ump) rw_assert(&(ump)->um_softdep->sd_fslock, \ RA_WLOCKED) #define BUF_AREC(bp) lockallowrecurse(&(bp)->b_lock) #define BUF_NOREC(bp) lockdisablerecurse(&(bp)->b_lock) /* * Worklist queue management. * These routines require that the lock be held. */ #ifndef /* NOT */ DEBUG #define WORKLIST_INSERT(head, item) do { \ (item)->wk_state |= ONWORKLIST; \ LIST_INSERT_HEAD(head, item, wk_list); \ } while (0) #define WORKLIST_REMOVE(item) do { \ (item)->wk_state &= ~ONWORKLIST; \ LIST_REMOVE(item, wk_list); \ } while (0) #define WORKLIST_INSERT_UNLOCKED WORKLIST_INSERT #define WORKLIST_REMOVE_UNLOCKED WORKLIST_REMOVE #else /* DEBUG */ static void worklist_insert(struct workhead *, struct worklist *, int); static void worklist_remove(struct worklist *, int); #define WORKLIST_INSERT(head, item) worklist_insert(head, item, 1) #define WORKLIST_INSERT_UNLOCKED(head, item) worklist_insert(head, item, 0) #define WORKLIST_REMOVE(item) worklist_remove(item, 1) #define WORKLIST_REMOVE_UNLOCKED(item) worklist_remove(item, 0) static void worklist_insert(head, item, locked) struct workhead *head; struct worklist *item; int locked; { if (locked) LOCK_OWNED(VFSTOUFS(item->wk_mp)); if (item->wk_state & ONWORKLIST) panic("worklist_insert: %p %s(0x%X) already on list", item, TYPENAME(item->wk_type), item->wk_state); item->wk_state |= ONWORKLIST; LIST_INSERT_HEAD(head, item, wk_list); } static void worklist_remove(item, locked) struct worklist *item; int locked; { if (locked) LOCK_OWNED(VFSTOUFS(item->wk_mp)); if ((item->wk_state & ONWORKLIST) == 0) panic("worklist_remove: %p %s(0x%X) not on list", item, TYPENAME(item->wk_type), item->wk_state); item->wk_state &= ~ONWORKLIST; LIST_REMOVE(item, wk_list); } #endif /* DEBUG */ /* * Merge two jsegdeps keeping only the oldest one as newer references * can't be discarded until after older references. */ static inline struct jsegdep * jsegdep_merge(struct jsegdep *one, struct jsegdep *two) { struct jsegdep *swp; if (two == NULL) return (one); if (one->jd_seg->js_seq > two->jd_seg->js_seq) { swp = one; one = two; two = swp; } WORKLIST_REMOVE(&two->jd_list); free_jsegdep(two); return (one); } /* * If two freedeps are compatible free one to reduce list size. */ static inline struct freedep * freedep_merge(struct freedep *one, struct freedep *two) { if (two == NULL) return (one); if (one->fd_freework == two->fd_freework) { WORKLIST_REMOVE(&two->fd_list); free_freedep(two); } return (one); } /* * Move journal work from one list to another. Duplicate freedeps and * jsegdeps are coalesced to keep the lists as small as possible. */ static void jwork_move(dst, src) struct workhead *dst; struct workhead *src; { struct freedep *freedep; struct jsegdep *jsegdep; struct worklist *wkn; struct worklist *wk; KASSERT(dst != src, ("jwork_move: dst == src")); freedep = NULL; jsegdep = NULL; LIST_FOREACH_SAFE(wk, dst, wk_list, wkn) { if (wk->wk_type == D_JSEGDEP) jsegdep = jsegdep_merge(WK_JSEGDEP(wk), jsegdep); else if (wk->wk_type == D_FREEDEP) freedep = freedep_merge(WK_FREEDEP(wk), freedep); } while ((wk = LIST_FIRST(src)) != NULL) { WORKLIST_REMOVE(wk); WORKLIST_INSERT(dst, wk); if (wk->wk_type == D_JSEGDEP) { jsegdep = jsegdep_merge(WK_JSEGDEP(wk), jsegdep); continue; } if (wk->wk_type == D_FREEDEP) freedep = freedep_merge(WK_FREEDEP(wk), freedep); } } static void jwork_insert(dst, jsegdep) struct workhead *dst; struct jsegdep *jsegdep; { struct jsegdep *jsegdepn; struct worklist *wk; LIST_FOREACH(wk, dst, wk_list) if (wk->wk_type == D_JSEGDEP) break; if (wk == NULL) { WORKLIST_INSERT(dst, &jsegdep->jd_list); return; } jsegdepn = WK_JSEGDEP(wk); if (jsegdep->jd_seg->js_seq < jsegdepn->jd_seg->js_seq) { WORKLIST_REMOVE(wk); free_jsegdep(jsegdepn); WORKLIST_INSERT(dst, &jsegdep->jd_list); } else free_jsegdep(jsegdep); } /* * Routines for tracking and managing workitems. */ static void workitem_free(struct worklist *, int); static void workitem_alloc(struct worklist *, int, struct mount *); static void workitem_reassign(struct worklist *, int); #define WORKITEM_FREE(item, type) \ workitem_free((struct worklist *)(item), (type)) #define WORKITEM_REASSIGN(item, type) \ workitem_reassign((struct worklist *)(item), (type)) static void workitem_free(item, type) struct worklist *item; int type; { struct ufsmount *ump; #ifdef DEBUG if (item->wk_state & ONWORKLIST) panic("workitem_free: %s(0x%X) still on list", TYPENAME(item->wk_type), item->wk_state); if (item->wk_type != type && type != D_NEWBLK) panic("workitem_free: type mismatch %s != %s", TYPENAME(item->wk_type), TYPENAME(type)); #endif if (item->wk_state & IOWAITING) wakeup(item); ump = VFSTOUFS(item->wk_mp); LOCK_OWNED(ump); KASSERT(ump->softdep_deps > 0, ("workitem_free: %s: softdep_deps going negative", ump->um_fs->fs_fsmnt)); if (--ump->softdep_deps == 0 && ump->softdep_req) wakeup(&ump->softdep_deps); KASSERT(dep_current[item->wk_type] > 0, ("workitem_free: %s: dep_current[%s] going negative", ump->um_fs->fs_fsmnt, TYPENAME(item->wk_type))); KASSERT(ump->softdep_curdeps[item->wk_type] > 0, ("workitem_free: %s: softdep_curdeps[%s] going negative", ump->um_fs->fs_fsmnt, TYPENAME(item->wk_type))); atomic_subtract_long(&dep_current[item->wk_type], 1); ump->softdep_curdeps[item->wk_type] -= 1; free(item, DtoM(type)); } static void workitem_alloc(item, type, mp) struct worklist *item; int type; struct mount *mp; { struct ufsmount *ump; item->wk_type = type; item->wk_mp = mp; item->wk_state = 0; ump = VFSTOUFS(mp); ACQUIRE_GBLLOCK(&lk); dep_current[type]++; if (dep_current[type] > dep_highuse[type]) dep_highuse[type] = dep_current[type]; dep_total[type]++; FREE_GBLLOCK(&lk); ACQUIRE_LOCK(ump); ump->softdep_curdeps[type] += 1; ump->softdep_deps++; ump->softdep_accdeps++; FREE_LOCK(ump); } static void workitem_reassign(item, newtype) struct worklist *item; int newtype; { struct ufsmount *ump; ump = VFSTOUFS(item->wk_mp); LOCK_OWNED(ump); KASSERT(ump->softdep_curdeps[item->wk_type] > 0, ("workitem_reassign: %s: softdep_curdeps[%s] going negative", VFSTOUFS(item->wk_mp)->um_fs->fs_fsmnt, TYPENAME(item->wk_type))); ump->softdep_curdeps[item->wk_type] -= 1; ump->softdep_curdeps[newtype] += 1; KASSERT(dep_current[item->wk_type] > 0, ("workitem_reassign: %s: dep_current[%s] going negative", VFSTOUFS(item->wk_mp)->um_fs->fs_fsmnt, TYPENAME(item->wk_type))); ACQUIRE_GBLLOCK(&lk); dep_current[newtype]++; dep_current[item->wk_type]--; if (dep_current[newtype] > dep_highuse[newtype]) dep_highuse[newtype] = dep_current[newtype]; dep_total[newtype]++; FREE_GBLLOCK(&lk); item->wk_type = newtype; } /* * Workitem queue management */ static int max_softdeps; /* maximum number of structs before slowdown */ static int tickdelay = 2; /* number of ticks to pause during slowdown */ static int proc_waiting; /* tracks whether we have a timeout posted */ static int *stat_countp; /* statistic to count in proc_waiting timeout */ static struct callout softdep_callout; static int req_clear_inodedeps; /* syncer process flush some inodedeps */ static int req_clear_remove; /* syncer process flush some freeblks */ static int softdep_flushcache = 0; /* Should we do BIO_FLUSH? */ /* * runtime statistics */ static int stat_flush_threads; /* number of softdep flushing threads */ static int stat_worklist_push; /* number of worklist cleanups */ static int stat_blk_limit_push; /* number of times block limit neared */ static int stat_ino_limit_push; /* number of times inode limit neared */ static int stat_blk_limit_hit; /* number of times block slowdown imposed */ static int stat_ino_limit_hit; /* number of times inode slowdown imposed */ static int stat_sync_limit_hit; /* number of synchronous slowdowns imposed */ static int stat_indir_blk_ptrs; /* bufs redirtied as indir ptrs not written */ static int stat_inode_bitmap; /* bufs redirtied as inode bitmap not written */ static int stat_direct_blk_ptrs;/* bufs redirtied as direct ptrs not written */ static int stat_dir_entry; /* bufs redirtied as dir entry cannot write */ static int stat_jaddref; /* bufs redirtied as ino bitmap can not write */ static int stat_jnewblk; /* bufs redirtied as blk bitmap can not write */ static int stat_journal_min; /* Times hit journal min threshold */ static int stat_journal_low; /* Times hit journal low threshold */ static int stat_journal_wait; /* Times blocked in jwait(). */ static int stat_jwait_filepage; /* Times blocked in jwait() for filepage. */ static int stat_jwait_freeblks; /* Times blocked in jwait() for freeblks. */ static int stat_jwait_inode; /* Times blocked in jwait() for inodes. */ static int stat_jwait_newblk; /* Times blocked in jwait() for newblks. */ static int stat_cleanup_high_delay; /* Maximum cleanup delay (in ticks) */ static int stat_cleanup_blkrequests; /* Number of block cleanup requests */ static int stat_cleanup_inorequests; /* Number of inode cleanup requests */ static int stat_cleanup_retries; /* Number of cleanups that needed to flush */ static int stat_cleanup_failures; /* Number of cleanup requests that failed */ static int stat_emptyjblocks; /* Number of potentially empty journal blocks */ SYSCTL_INT(_debug_softdep, OID_AUTO, max_softdeps, CTLFLAG_RW, &max_softdeps, 0, ""); SYSCTL_INT(_debug_softdep, OID_AUTO, tickdelay, CTLFLAG_RW, &tickdelay, 0, ""); SYSCTL_INT(_debug_softdep, OID_AUTO, flush_threads, CTLFLAG_RD, &stat_flush_threads, 0, ""); SYSCTL_INT(_debug_softdep, OID_AUTO, worklist_push, CTLFLAG_RW, &stat_worklist_push, 0,""); SYSCTL_INT(_debug_softdep, OID_AUTO, blk_limit_push, CTLFLAG_RW, &stat_blk_limit_push, 0,""); SYSCTL_INT(_debug_softdep, OID_AUTO, ino_limit_push, CTLFLAG_RW, &stat_ino_limit_push, 0,""); SYSCTL_INT(_debug_softdep, OID_AUTO, blk_limit_hit, CTLFLAG_RW, &stat_blk_limit_hit, 0, ""); SYSCTL_INT(_debug_softdep, OID_AUTO, ino_limit_hit, CTLFLAG_RW, &stat_ino_limit_hit, 0, ""); SYSCTL_INT(_debug_softdep, OID_AUTO, sync_limit_hit, CTLFLAG_RW, &stat_sync_limit_hit, 0, ""); SYSCTL_INT(_debug_softdep, OID_AUTO, indir_blk_ptrs, CTLFLAG_RW, &stat_indir_blk_ptrs, 0, ""); SYSCTL_INT(_debug_softdep, OID_AUTO, inode_bitmap, CTLFLAG_RW, &stat_inode_bitmap, 0, ""); SYSCTL_INT(_debug_softdep, OID_AUTO, direct_blk_ptrs, CTLFLAG_RW, &stat_direct_blk_ptrs, 0, ""); SYSCTL_INT(_debug_softdep, OID_AUTO, dir_entry, CTLFLAG_RW, &stat_dir_entry, 0, ""); SYSCTL_INT(_debug_softdep, OID_AUTO, jaddref_rollback, CTLFLAG_RW, &stat_jaddref, 0, ""); SYSCTL_INT(_debug_softdep, OID_AUTO, jnewblk_rollback, CTLFLAG_RW, &stat_jnewblk, 0, ""); SYSCTL_INT(_debug_softdep, OID_AUTO, journal_low, CTLFLAG_RW, &stat_journal_low, 0, ""); SYSCTL_INT(_debug_softdep, OID_AUTO, journal_min, CTLFLAG_RW, &stat_journal_min, 0, ""); SYSCTL_INT(_debug_softdep, OID_AUTO, journal_wait, CTLFLAG_RW, &stat_journal_wait, 0, ""); SYSCTL_INT(_debug_softdep, OID_AUTO, jwait_filepage, CTLFLAG_RW, &stat_jwait_filepage, 0, ""); SYSCTL_INT(_debug_softdep, OID_AUTO, jwait_freeblks, CTLFLAG_RW, &stat_jwait_freeblks, 0, ""); SYSCTL_INT(_debug_softdep, OID_AUTO, jwait_inode, CTLFLAG_RW, &stat_jwait_inode, 0, ""); SYSCTL_INT(_debug_softdep, OID_AUTO, jwait_newblk, CTLFLAG_RW, &stat_jwait_newblk, 0, ""); SYSCTL_INT(_debug_softdep, OID_AUTO, cleanup_blkrequests, CTLFLAG_RW, &stat_cleanup_blkrequests, 0, ""); SYSCTL_INT(_debug_softdep, OID_AUTO, cleanup_inorequests, CTLFLAG_RW, &stat_cleanup_inorequests, 0, ""); SYSCTL_INT(_debug_softdep, OID_AUTO, cleanup_high_delay, CTLFLAG_RW, &stat_cleanup_high_delay, 0, ""); SYSCTL_INT(_debug_softdep, OID_AUTO, cleanup_retries, CTLFLAG_RW, &stat_cleanup_retries, 0, ""); SYSCTL_INT(_debug_softdep, OID_AUTO, cleanup_failures, CTLFLAG_RW, &stat_cleanup_failures, 0, ""); SYSCTL_INT(_debug_softdep, OID_AUTO, flushcache, CTLFLAG_RW, &softdep_flushcache, 0, ""); SYSCTL_INT(_debug_softdep, OID_AUTO, emptyjblocks, CTLFLAG_RD, &stat_emptyjblocks, 0, ""); SYSCTL_DECL(_vfs_ffs); /* Whether to recompute the summary at mount time */ static int compute_summary_at_mount = 0; SYSCTL_INT(_vfs_ffs, OID_AUTO, compute_summary_at_mount, CTLFLAG_RW, &compute_summary_at_mount, 0, "Recompute summary at mount"); static int print_threads = 0; SYSCTL_INT(_debug_softdep, OID_AUTO, print_threads, CTLFLAG_RW, &print_threads, 0, "Notify flusher thread start/stop"); /* List of all filesystems mounted with soft updates */ static TAILQ_HEAD(, mount_softdeps) softdepmounts; /* * This function cleans the worklist for a filesystem. * Each filesystem running with soft dependencies gets its own * thread to run in this function. The thread is started up in * softdep_mount and shutdown in softdep_unmount. They show up * as part of the kernel "bufdaemon" process whose process * entry is available in bufdaemonproc. */ static int searchfailed; extern struct proc *bufdaemonproc; static void softdep_flush(addr) void *addr; { struct mount *mp; struct thread *td; struct ufsmount *ump; td = curthread; td->td_pflags |= TDP_NORUNNINGBUF; mp = (struct mount *)addr; ump = VFSTOUFS(mp); atomic_add_int(&stat_flush_threads, 1); ACQUIRE_LOCK(ump); ump->softdep_flags &= ~FLUSH_STARTING; wakeup(&ump->softdep_flushtd); FREE_LOCK(ump); if (print_threads) { if (stat_flush_threads == 1) printf("Running %s at pid %d\n", bufdaemonproc->p_comm, bufdaemonproc->p_pid); printf("Start thread %s\n", td->td_name); } for (;;) { while (softdep_process_worklist(mp, 0) > 0 || (MOUNTEDSUJ(mp) && VFSTOUFS(mp)->softdep_jblocks->jb_suspended)) kthread_suspend_check(); ACQUIRE_LOCK(ump); if ((ump->softdep_flags & (FLUSH_CLEANUP | FLUSH_EXIT)) == 0) msleep(&ump->softdep_flushtd, LOCK_PTR(ump), PVM, "sdflush", hz / 2); ump->softdep_flags &= ~FLUSH_CLEANUP; /* * Check to see if we are done and need to exit. */ if ((ump->softdep_flags & FLUSH_EXIT) == 0) { FREE_LOCK(ump); continue; } ump->softdep_flags &= ~FLUSH_EXIT; FREE_LOCK(ump); wakeup(&ump->softdep_flags); if (print_threads) printf("Stop thread %s: searchfailed %d, did cleanups %d\n", td->td_name, searchfailed, ump->um_softdep->sd_cleanups); atomic_subtract_int(&stat_flush_threads, 1); kthread_exit(); panic("kthread_exit failed\n"); } } static void worklist_speedup(mp) struct mount *mp; { struct ufsmount *ump; ump = VFSTOUFS(mp); LOCK_OWNED(ump); if ((ump->softdep_flags & (FLUSH_CLEANUP | FLUSH_EXIT)) == 0) ump->softdep_flags |= FLUSH_CLEANUP; wakeup(&ump->softdep_flushtd); } static int softdep_speedup(ump) struct ufsmount *ump; { struct ufsmount *altump; struct mount_softdeps *sdp; LOCK_OWNED(ump); worklist_speedup(ump->um_mountp); bd_speedup(); /* * If we have global shortages, then we need other * filesystems to help with the cleanup. Here we wakeup a * flusher thread for a filesystem that is over its fair * share of resources. */ if (req_clear_inodedeps || req_clear_remove) { ACQUIRE_GBLLOCK(&lk); TAILQ_FOREACH(sdp, &softdepmounts, sd_next) { if ((altump = sdp->sd_ump) == ump) continue; if (((req_clear_inodedeps && altump->softdep_curdeps[D_INODEDEP] > max_softdeps / stat_flush_threads) || (req_clear_remove && altump->softdep_curdeps[D_DIRREM] > (max_softdeps / 2) / stat_flush_threads)) && TRY_ACQUIRE_LOCK(altump)) break; } if (sdp == NULL) { searchfailed++; FREE_GBLLOCK(&lk); } else { /* * Move to the end of the list so we pick a * different one on out next try. */ TAILQ_REMOVE(&softdepmounts, sdp, sd_next); TAILQ_INSERT_TAIL(&softdepmounts, sdp, sd_next); FREE_GBLLOCK(&lk); if ((altump->softdep_flags & (FLUSH_CLEANUP | FLUSH_EXIT)) == 0) altump->softdep_flags |= FLUSH_CLEANUP; altump->um_softdep->sd_cleanups++; wakeup(&altump->softdep_flushtd); FREE_LOCK(altump); } } return (speedup_syncer()); } /* * Add an item to the end of the work queue. * This routine requires that the lock be held. * This is the only routine that adds items to the list. * The following routine is the only one that removes items * and does so in order from first to last. */ #define WK_HEAD 0x0001 /* Add to HEAD. */ #define WK_NODELAY 0x0002 /* Process immediately. */ static void add_to_worklist(wk, flags) struct worklist *wk; int flags; { struct ufsmount *ump; ump = VFSTOUFS(wk->wk_mp); LOCK_OWNED(ump); if (wk->wk_state & ONWORKLIST) panic("add_to_worklist: %s(0x%X) already on list", TYPENAME(wk->wk_type), wk->wk_state); wk->wk_state |= ONWORKLIST; if (ump->softdep_on_worklist == 0) { LIST_INSERT_HEAD(&ump->softdep_workitem_pending, wk, wk_list); ump->softdep_worklist_tail = wk; } else if (flags & WK_HEAD) { LIST_INSERT_HEAD(&ump->softdep_workitem_pending, wk, wk_list); } else { LIST_INSERT_AFTER(ump->softdep_worklist_tail, wk, wk_list); ump->softdep_worklist_tail = wk; } ump->softdep_on_worklist += 1; if (flags & WK_NODELAY) worklist_speedup(wk->wk_mp); } /* * Remove the item to be processed. If we are removing the last * item on the list, we need to recalculate the tail pointer. */ static void remove_from_worklist(wk) struct worklist *wk; { struct ufsmount *ump; ump = VFSTOUFS(wk->wk_mp); WORKLIST_REMOVE(wk); if (ump->softdep_worklist_tail == wk) ump->softdep_worklist_tail = (struct worklist *)wk->wk_list.le_prev; ump->softdep_on_worklist -= 1; } static void wake_worklist(wk) struct worklist *wk; { if (wk->wk_state & IOWAITING) { wk->wk_state &= ~IOWAITING; wakeup(wk); } } static void wait_worklist(wk, wmesg) struct worklist *wk; char *wmesg; { struct ufsmount *ump; ump = VFSTOUFS(wk->wk_mp); wk->wk_state |= IOWAITING; msleep(wk, LOCK_PTR(ump), PVM, wmesg, 0); } /* * Process that runs once per second to handle items in the background queue. * * Note that we ensure that everything is done in the order in which they * appear in the queue. The code below depends on this property to ensure * that blocks of a file are freed before the inode itself is freed. This * ordering ensures that no new triples will be generated * until all the old ones have been purged from the dependency lists. */ static int softdep_process_worklist(mp, full) struct mount *mp; int full; { int cnt, matchcnt; struct ufsmount *ump; long starttime; KASSERT(mp != NULL, ("softdep_process_worklist: NULL mp")); if (MOUNTEDSOFTDEP(mp) == 0) return (0); matchcnt = 0; ump = VFSTOUFS(mp); ACQUIRE_LOCK(ump); starttime = time_second; softdep_process_journal(mp, NULL, full ? MNT_WAIT : 0); check_clear_deps(mp); while (ump->softdep_on_worklist > 0) { if ((cnt = process_worklist_item(mp, 10, LK_NOWAIT)) == 0) break; else matchcnt += cnt; check_clear_deps(mp); /* * We do not generally want to stop for buffer space, but if * we are really being a buffer hog, we will stop and wait. */ if (should_yield()) { FREE_LOCK(ump); kern_yield(PRI_USER); bwillwrite(); ACQUIRE_LOCK(ump); } /* * Never allow processing to run for more than one * second. This gives the syncer thread the opportunity * to pause if appropriate. */ if (!full && starttime != time_second) break; } if (full == 0) journal_unsuspend(ump); FREE_LOCK(ump); return (matchcnt); } /* * Process all removes associated with a vnode if we are running out of * journal space. Any other process which attempts to flush these will * be unable as we have the vnodes locked. */ static void process_removes(vp) struct vnode *vp; { struct inodedep *inodedep; struct dirrem *dirrem; struct ufsmount *ump; struct mount *mp; ino_t inum; mp = vp->v_mount; ump = VFSTOUFS(mp); LOCK_OWNED(ump); inum = VTOI(vp)->i_number; for (;;) { top: if (inodedep_lookup(mp, inum, 0, &inodedep) == 0) return; LIST_FOREACH(dirrem, &inodedep->id_dirremhd, dm_inonext) { /* * If another thread is trying to lock this vnode * it will fail but we must wait for it to do so * before we can proceed. */ if (dirrem->dm_state & INPROGRESS) { wait_worklist(&dirrem->dm_list, "pwrwait"); goto top; } if ((dirrem->dm_state & (COMPLETE | ONWORKLIST)) == (COMPLETE | ONWORKLIST)) break; } if (dirrem == NULL) return; remove_from_worklist(&dirrem->dm_list); FREE_LOCK(ump); if (vn_start_secondary_write(NULL, &mp, V_NOWAIT)) panic("process_removes: suspended filesystem"); handle_workitem_remove(dirrem, 0); vn_finished_secondary_write(mp); ACQUIRE_LOCK(ump); } } /* * Process all truncations associated with a vnode if we are running out * of journal space. This is called when the vnode lock is already held * and no other process can clear the truncation. This function returns * a value greater than zero if it did any work. */ static void process_truncates(vp) struct vnode *vp; { struct inodedep *inodedep; struct freeblks *freeblks; struct ufsmount *ump; struct mount *mp; ino_t inum; int cgwait; mp = vp->v_mount; ump = VFSTOUFS(mp); LOCK_OWNED(ump); inum = VTOI(vp)->i_number; for (;;) { if (inodedep_lookup(mp, inum, 0, &inodedep) == 0) return; cgwait = 0; TAILQ_FOREACH(freeblks, &inodedep->id_freeblklst, fb_next) { /* Journal entries not yet written. */ if (!LIST_EMPTY(&freeblks->fb_jblkdephd)) { jwait(&LIST_FIRST( &freeblks->fb_jblkdephd)->jb_list, MNT_WAIT); break; } /* Another thread is executing this item. */ if (freeblks->fb_state & INPROGRESS) { wait_worklist(&freeblks->fb_list, "ptrwait"); break; } /* Freeblks is waiting on a inode write. */ if ((freeblks->fb_state & COMPLETE) == 0) { FREE_LOCK(ump); ffs_update(vp, 1); ACQUIRE_LOCK(ump); break; } if ((freeblks->fb_state & (ALLCOMPLETE | ONWORKLIST)) == (ALLCOMPLETE | ONWORKLIST)) { remove_from_worklist(&freeblks->fb_list); freeblks->fb_state |= INPROGRESS; FREE_LOCK(ump); if (vn_start_secondary_write(NULL, &mp, V_NOWAIT)) panic("process_truncates: " "suspended filesystem"); handle_workitem_freeblocks(freeblks, 0); vn_finished_secondary_write(mp); ACQUIRE_LOCK(ump); break; } if (freeblks->fb_cgwait) cgwait++; } if (cgwait) { FREE_LOCK(ump); sync_cgs(mp, MNT_WAIT); ffs_sync_snap(mp, MNT_WAIT); ACQUIRE_LOCK(ump); continue; } if (freeblks == NULL) break; } return; } /* * Process one item on the worklist. */ static int process_worklist_item(mp, target, flags) struct mount *mp; int target; int flags; { struct worklist sentinel; struct worklist *wk; struct ufsmount *ump; int matchcnt; int error; KASSERT(mp != NULL, ("process_worklist_item: NULL mp")); /* * If we are being called because of a process doing a * copy-on-write, then it is not safe to write as we may * recurse into the copy-on-write routine. */ if (curthread->td_pflags & TDP_COWINPROGRESS) return (-1); PHOLD(curproc); /* Don't let the stack go away. */ ump = VFSTOUFS(mp); LOCK_OWNED(ump); matchcnt = 0; sentinel.wk_mp = NULL; sentinel.wk_type = D_SENTINEL; LIST_INSERT_HEAD(&ump->softdep_workitem_pending, &sentinel, wk_list); for (wk = LIST_NEXT(&sentinel, wk_list); wk != NULL; wk = LIST_NEXT(&sentinel, wk_list)) { if (wk->wk_type == D_SENTINEL) { LIST_REMOVE(&sentinel, wk_list); LIST_INSERT_AFTER(wk, &sentinel, wk_list); continue; } if (wk->wk_state & INPROGRESS) panic("process_worklist_item: %p already in progress.", wk); wk->wk_state |= INPROGRESS; remove_from_worklist(wk); FREE_LOCK(ump); if (vn_start_secondary_write(NULL, &mp, V_NOWAIT)) panic("process_worklist_item: suspended filesystem"); switch (wk->wk_type) { case D_DIRREM: /* removal of a directory entry */ error = handle_workitem_remove(WK_DIRREM(wk), flags); break; case D_FREEBLKS: /* releasing blocks and/or fragments from a file */ error = handle_workitem_freeblocks(WK_FREEBLKS(wk), flags); break; case D_FREEFRAG: /* releasing a fragment when replaced as a file grows */ handle_workitem_freefrag(WK_FREEFRAG(wk)); error = 0; break; case D_FREEFILE: /* releasing an inode when its link count drops to 0 */ handle_workitem_freefile(WK_FREEFILE(wk)); error = 0; break; default: panic("%s_process_worklist: Unknown type %s", "softdep", TYPENAME(wk->wk_type)); /* NOTREACHED */ } vn_finished_secondary_write(mp); ACQUIRE_LOCK(ump); if (error == 0) { if (++matchcnt == target) break; continue; } /* * We have to retry the worklist item later. Wake up any * waiters who may be able to complete it immediately and * add the item back to the head so we don't try to execute * it again. */ wk->wk_state &= ~INPROGRESS; wake_worklist(wk); add_to_worklist(wk, WK_HEAD); } LIST_REMOVE(&sentinel, wk_list); /* Sentinal could've become the tail from remove_from_worklist. */ if (ump->softdep_worklist_tail == &sentinel) ump->softdep_worklist_tail = (struct worklist *)sentinel.wk_list.le_prev; PRELE(curproc); return (matchcnt); } /* * Move dependencies from one buffer to another. */ int softdep_move_dependencies(oldbp, newbp) struct buf *oldbp; struct buf *newbp; { struct worklist *wk, *wktail; struct ufsmount *ump; int dirty; if ((wk = LIST_FIRST(&oldbp->b_dep)) == NULL) return (0); KASSERT(MOUNTEDSOFTDEP(wk->wk_mp) != 0, ("softdep_move_dependencies called on non-softdep filesystem")); dirty = 0; wktail = NULL; ump = VFSTOUFS(wk->wk_mp); ACQUIRE_LOCK(ump); while ((wk = LIST_FIRST(&oldbp->b_dep)) != NULL) { LIST_REMOVE(wk, wk_list); if (wk->wk_type == D_BMSAFEMAP && bmsafemap_backgroundwrite(WK_BMSAFEMAP(wk), newbp)) dirty = 1; if (wktail == 0) LIST_INSERT_HEAD(&newbp->b_dep, wk, wk_list); else LIST_INSERT_AFTER(wktail, wk, wk_list); wktail = wk; } FREE_LOCK(ump); return (dirty); } /* * Purge the work list of all items associated with a particular mount point. */ int softdep_flushworklist(oldmnt, countp, td) struct mount *oldmnt; int *countp; struct thread *td; { struct vnode *devvp; struct ufsmount *ump; int count, error; /* * Alternately flush the block device associated with the mount * point and process any dependencies that the flushing * creates. We continue until no more worklist dependencies * are found. */ *countp = 0; error = 0; ump = VFSTOUFS(oldmnt); devvp = ump->um_devvp; while ((count = softdep_process_worklist(oldmnt, 1)) > 0) { *countp += count; vn_lock(devvp, LK_EXCLUSIVE | LK_RETRY); error = VOP_FSYNC(devvp, MNT_WAIT, td); VOP_UNLOCK(devvp, 0); if (error != 0) break; } return (error); } #define SU_WAITIDLE_RETRIES 20 static int softdep_waitidle(struct mount *mp, int flags __unused) { struct ufsmount *ump; struct vnode *devvp; struct thread *td; int error, i; ump = VFSTOUFS(mp); devvp = ump->um_devvp; td = curthread; error = 0; ACQUIRE_LOCK(ump); for (i = 0; i < SU_WAITIDLE_RETRIES && ump->softdep_deps != 0; i++) { ump->softdep_req = 1; KASSERT((flags & FORCECLOSE) == 0 || ump->softdep_on_worklist == 0, ("softdep_waitidle: work added after flush")); msleep(&ump->softdep_deps, LOCK_PTR(ump), PVM | PDROP, "softdeps", 10 * hz); vn_lock(devvp, LK_EXCLUSIVE | LK_RETRY); error = VOP_FSYNC(devvp, MNT_WAIT, td); VOP_UNLOCK(devvp, 0); if (error != 0) break; ACQUIRE_LOCK(ump); } ump->softdep_req = 0; if (i == SU_WAITIDLE_RETRIES && error == 0 && ump->softdep_deps != 0) { error = EBUSY; printf("softdep_waitidle: Failed to flush worklist for %p\n", mp); } FREE_LOCK(ump); return (error); } /* * Flush all vnodes and worklist items associated with a specified mount point. */ int softdep_flushfiles(oldmnt, flags, td) struct mount *oldmnt; int flags; struct thread *td; { #ifdef QUOTA struct ufsmount *ump; int i; #endif int error, early, depcount, loopcnt, retry_flush_count, retry; int morework; KASSERT(MOUNTEDSOFTDEP(oldmnt) != 0, ("softdep_flushfiles called on non-softdep filesystem")); loopcnt = 10; retry_flush_count = 3; retry_flush: error = 0; /* * Alternately flush the vnodes associated with the mount * point and process any dependencies that the flushing * creates. In theory, this loop can happen at most twice, * but we give it a few extra just to be sure. */ for (; loopcnt > 0; loopcnt--) { /* * Do another flush in case any vnodes were brought in * as part of the cleanup operations. */ early = retry_flush_count == 1 || (oldmnt->mnt_kern_flag & MNTK_UNMOUNT) == 0 ? 0 : EARLYFLUSH; if ((error = ffs_flushfiles(oldmnt, flags | early, td)) != 0) break; if ((error = softdep_flushworklist(oldmnt, &depcount, td)) != 0 || depcount == 0) break; } /* * If we are unmounting then it is an error to fail. If we * are simply trying to downgrade to read-only, then filesystem * activity can keep us busy forever, so we just fail with EBUSY. */ if (loopcnt == 0) { if (oldmnt->mnt_kern_flag & MNTK_UNMOUNT) panic("softdep_flushfiles: looping"); error = EBUSY; } if (!error) error = softdep_waitidle(oldmnt, flags); if (!error) { if (oldmnt->mnt_kern_flag & MNTK_UNMOUNT) { retry = 0; MNT_ILOCK(oldmnt); KASSERT((oldmnt->mnt_kern_flag & MNTK_NOINSMNTQ) != 0, ("softdep_flushfiles: !MNTK_NOINSMNTQ")); morework = oldmnt->mnt_nvnodelistsize > 0; #ifdef QUOTA ump = VFSTOUFS(oldmnt); UFS_LOCK(ump); for (i = 0; i < MAXQUOTAS; i++) { if (ump->um_quotas[i] != NULLVP) morework = 1; } UFS_UNLOCK(ump); #endif if (morework) { if (--retry_flush_count > 0) { retry = 1; loopcnt = 3; } else error = EBUSY; } MNT_IUNLOCK(oldmnt); if (retry) goto retry_flush; } } return (error); } /* * Structure hashing. * * There are four types of structures that can be looked up: * 1) pagedep structures identified by mount point, inode number, * and logical block. * 2) inodedep structures identified by mount point and inode number. * 3) newblk structures identified by mount point and * physical block number. * 4) bmsafemap structures identified by mount point and * cylinder group number. * * The "pagedep" and "inodedep" dependency structures are hashed * separately from the file blocks and inodes to which they correspond. * This separation helps when the in-memory copy of an inode or * file block must be replaced. It also obviates the need to access * an inode or file page when simply updating (or de-allocating) * dependency structures. Lookup of newblk structures is needed to * find newly allocated blocks when trying to associate them with * their allocdirect or allocindir structure. * * The lookup routines optionally create and hash a new instance when * an existing entry is not found. The bmsafemap lookup routine always * allocates a new structure if an existing one is not found. */ #define DEPALLOC 0x0001 /* allocate structure if lookup fails */ /* * Structures and routines associated with pagedep caching. */ #define PAGEDEP_HASH(ump, inum, lbn) \ (&(ump)->pagedep_hashtbl[((inum) + (lbn)) & (ump)->pagedep_hash_size]) static int pagedep_find(pagedephd, ino, lbn, pagedeppp) struct pagedep_hashhead *pagedephd; ino_t ino; ufs_lbn_t lbn; struct pagedep **pagedeppp; { struct pagedep *pagedep; LIST_FOREACH(pagedep, pagedephd, pd_hash) { if (ino == pagedep->pd_ino && lbn == pagedep->pd_lbn) { *pagedeppp = pagedep; return (1); } } *pagedeppp = NULL; return (0); } /* * Look up a pagedep. Return 1 if found, 0 otherwise. * If not found, allocate if DEPALLOC flag is passed. * Found or allocated entry is returned in pagedeppp. * This routine must be called with splbio interrupts blocked. */ static int pagedep_lookup(mp, bp, ino, lbn, flags, pagedeppp) struct mount *mp; struct buf *bp; ino_t ino; ufs_lbn_t lbn; int flags; struct pagedep **pagedeppp; { struct pagedep *pagedep; struct pagedep_hashhead *pagedephd; struct worklist *wk; struct ufsmount *ump; int ret; int i; ump = VFSTOUFS(mp); LOCK_OWNED(ump); if (bp) { LIST_FOREACH(wk, &bp->b_dep, wk_list) { if (wk->wk_type == D_PAGEDEP) { *pagedeppp = WK_PAGEDEP(wk); return (1); } } } pagedephd = PAGEDEP_HASH(ump, ino, lbn); ret = pagedep_find(pagedephd, ino, lbn, pagedeppp); if (ret) { if (((*pagedeppp)->pd_state & ONWORKLIST) == 0 && bp) WORKLIST_INSERT(&bp->b_dep, &(*pagedeppp)->pd_list); return (1); } if ((flags & DEPALLOC) == 0) return (0); FREE_LOCK(ump); pagedep = malloc(sizeof(struct pagedep), M_PAGEDEP, M_SOFTDEP_FLAGS|M_ZERO); workitem_alloc(&pagedep->pd_list, D_PAGEDEP, mp); ACQUIRE_LOCK(ump); ret = pagedep_find(pagedephd, ino, lbn, pagedeppp); if (*pagedeppp) { /* * This should never happen since we only create pagedeps * with the vnode lock held. Could be an assert. */ WORKITEM_FREE(pagedep, D_PAGEDEP); return (ret); } pagedep->pd_ino = ino; pagedep->pd_lbn = lbn; LIST_INIT(&pagedep->pd_dirremhd); LIST_INIT(&pagedep->pd_pendinghd); for (i = 0; i < DAHASHSZ; i++) LIST_INIT(&pagedep->pd_diraddhd[i]); LIST_INSERT_HEAD(pagedephd, pagedep, pd_hash); WORKLIST_INSERT(&bp->b_dep, &pagedep->pd_list); *pagedeppp = pagedep; return (0); } /* * Structures and routines associated with inodedep caching. */ #define INODEDEP_HASH(ump, inum) \ (&(ump)->inodedep_hashtbl[(inum) & (ump)->inodedep_hash_size]) static int inodedep_find(inodedephd, inum, inodedeppp) struct inodedep_hashhead *inodedephd; ino_t inum; struct inodedep **inodedeppp; { struct inodedep *inodedep; LIST_FOREACH(inodedep, inodedephd, id_hash) if (inum == inodedep->id_ino) break; if (inodedep) { *inodedeppp = inodedep; return (1); } *inodedeppp = NULL; return (0); } /* * Look up an inodedep. Return 1 if found, 0 if not found. * If not found, allocate if DEPALLOC flag is passed. * Found or allocated entry is returned in inodedeppp. * This routine must be called with splbio interrupts blocked. */ static int inodedep_lookup(mp, inum, flags, inodedeppp) struct mount *mp; ino_t inum; int flags; struct inodedep **inodedeppp; { struct inodedep *inodedep; struct inodedep_hashhead *inodedephd; struct ufsmount *ump; struct fs *fs; ump = VFSTOUFS(mp); LOCK_OWNED(ump); fs = ump->um_fs; inodedephd = INODEDEP_HASH(ump, inum); if (inodedep_find(inodedephd, inum, inodedeppp)) return (1); if ((flags & DEPALLOC) == 0) return (0); /* * If the system is over its limit and our filesystem is * responsible for more than our share of that usage and * we are not in a rush, request some inodedep cleanup. */ if (softdep_excess_inodes(ump)) schedule_cleanup(mp); else FREE_LOCK(ump); inodedep = malloc(sizeof(struct inodedep), M_INODEDEP, M_SOFTDEP_FLAGS); workitem_alloc(&inodedep->id_list, D_INODEDEP, mp); ACQUIRE_LOCK(ump); if (inodedep_find(inodedephd, inum, inodedeppp)) { WORKITEM_FREE(inodedep, D_INODEDEP); return (1); } inodedep->id_fs = fs; inodedep->id_ino = inum; inodedep->id_state = ALLCOMPLETE; inodedep->id_nlinkdelta = 0; inodedep->id_savedino1 = NULL; inodedep->id_savedsize = -1; inodedep->id_savedextsize = -1; inodedep->id_savednlink = -1; inodedep->id_bmsafemap = NULL; inodedep->id_mkdiradd = NULL; LIST_INIT(&inodedep->id_dirremhd); LIST_INIT(&inodedep->id_pendinghd); LIST_INIT(&inodedep->id_inowait); LIST_INIT(&inodedep->id_bufwait); TAILQ_INIT(&inodedep->id_inoreflst); TAILQ_INIT(&inodedep->id_inoupdt); TAILQ_INIT(&inodedep->id_newinoupdt); TAILQ_INIT(&inodedep->id_extupdt); TAILQ_INIT(&inodedep->id_newextupdt); TAILQ_INIT(&inodedep->id_freeblklst); LIST_INSERT_HEAD(inodedephd, inodedep, id_hash); *inodedeppp = inodedep; return (0); } /* * Structures and routines associated with newblk caching. */ #define NEWBLK_HASH(ump, inum) \ (&(ump)->newblk_hashtbl[(inum) & (ump)->newblk_hash_size]) static int newblk_find(newblkhd, newblkno, flags, newblkpp) struct newblk_hashhead *newblkhd; ufs2_daddr_t newblkno; int flags; struct newblk **newblkpp; { struct newblk *newblk; LIST_FOREACH(newblk, newblkhd, nb_hash) { if (newblkno != newblk->nb_newblkno) continue; /* * If we're creating a new dependency don't match those that * have already been converted to allocdirects. This is for * a frag extend. */ if ((flags & DEPALLOC) && newblk->nb_list.wk_type != D_NEWBLK) continue; break; } if (newblk) { *newblkpp = newblk; return (1); } *newblkpp = NULL; return (0); } /* * Look up a newblk. Return 1 if found, 0 if not found. * If not found, allocate if DEPALLOC flag is passed. * Found or allocated entry is returned in newblkpp. */ static int newblk_lookup(mp, newblkno, flags, newblkpp) struct mount *mp; ufs2_daddr_t newblkno; int flags; struct newblk **newblkpp; { struct newblk *newblk; struct newblk_hashhead *newblkhd; struct ufsmount *ump; ump = VFSTOUFS(mp); LOCK_OWNED(ump); newblkhd = NEWBLK_HASH(ump, newblkno); if (newblk_find(newblkhd, newblkno, flags, newblkpp)) return (1); if ((flags & DEPALLOC) == 0) return (0); FREE_LOCK(ump); newblk = malloc(sizeof(union allblk), M_NEWBLK, M_SOFTDEP_FLAGS | M_ZERO); workitem_alloc(&newblk->nb_list, D_NEWBLK, mp); ACQUIRE_LOCK(ump); if (newblk_find(newblkhd, newblkno, flags, newblkpp)) { WORKITEM_FREE(newblk, D_NEWBLK); return (1); } newblk->nb_freefrag = NULL; LIST_INIT(&newblk->nb_indirdeps); LIST_INIT(&newblk->nb_newdirblk); LIST_INIT(&newblk->nb_jwork); newblk->nb_state = ATTACHED; newblk->nb_newblkno = newblkno; LIST_INSERT_HEAD(newblkhd, newblk, nb_hash); *newblkpp = newblk; return (0); } /* * Structures and routines associated with freed indirect block caching. */ #define INDIR_HASH(ump, blkno) \ (&(ump)->indir_hashtbl[(blkno) & (ump)->indir_hash_size]) /* * Lookup an indirect block in the indir hash table. The freework is * removed and potentially freed. The caller must do a blocking journal * write before writing to the blkno. */ static int indirblk_lookup(mp, blkno) struct mount *mp; ufs2_daddr_t blkno; { struct freework *freework; struct indir_hashhead *wkhd; struct ufsmount *ump; ump = VFSTOUFS(mp); wkhd = INDIR_HASH(ump, blkno); TAILQ_FOREACH(freework, wkhd, fw_next) { if (freework->fw_blkno != blkno) continue; indirblk_remove(freework); return (1); } return (0); } /* * Insert an indirect block represented by freework into the indirblk * hash table so that it may prevent the block from being re-used prior * to the journal being written. */ static void indirblk_insert(freework) struct freework *freework; { struct jblocks *jblocks; struct jseg *jseg; struct ufsmount *ump; ump = VFSTOUFS(freework->fw_list.wk_mp); jblocks = ump->softdep_jblocks; jseg = TAILQ_LAST(&jblocks->jb_segs, jseglst); if (jseg == NULL) return; LIST_INSERT_HEAD(&jseg->js_indirs, freework, fw_segs); TAILQ_INSERT_HEAD(INDIR_HASH(ump, freework->fw_blkno), freework, fw_next); freework->fw_state &= ~DEPCOMPLETE; } static void indirblk_remove(freework) struct freework *freework; { struct ufsmount *ump; ump = VFSTOUFS(freework->fw_list.wk_mp); LIST_REMOVE(freework, fw_segs); TAILQ_REMOVE(INDIR_HASH(ump, freework->fw_blkno), freework, fw_next); freework->fw_state |= DEPCOMPLETE; if ((freework->fw_state & ALLCOMPLETE) == ALLCOMPLETE) WORKITEM_FREE(freework, D_FREEWORK); } /* * Executed during filesystem system initialization before * mounting any filesystems. */ void softdep_initialize() { TAILQ_INIT(&softdepmounts); max_softdeps = desiredvnodes * 4; /* initialise bioops hack */ bioops.io_start = softdep_disk_io_initiation; bioops.io_complete = softdep_disk_write_complete; bioops.io_deallocate = softdep_deallocate_dependencies; bioops.io_countdeps = softdep_count_dependencies; softdep_ast_cleanup = softdep_ast_cleanup_proc; /* Initialize the callout with an mtx. */ callout_init_mtx(&softdep_callout, &lk, 0); } /* * Executed after all filesystems have been unmounted during * filesystem module unload. */ void softdep_uninitialize() { /* clear bioops hack */ bioops.io_start = NULL; bioops.io_complete = NULL; bioops.io_deallocate = NULL; bioops.io_countdeps = NULL; softdep_ast_cleanup = NULL; callout_drain(&softdep_callout); } /* * Called at mount time to notify the dependency code that a * filesystem wishes to use it. */ int softdep_mount(devvp, mp, fs, cred) struct vnode *devvp; struct mount *mp; struct fs *fs; struct ucred *cred; { struct csum_total cstotal; struct mount_softdeps *sdp; struct ufsmount *ump; struct cg *cgp; struct buf *bp; int i, error, cyl; sdp = malloc(sizeof(struct mount_softdeps), M_MOUNTDATA, M_WAITOK | M_ZERO); MNT_ILOCK(mp); mp->mnt_flag = (mp->mnt_flag & ~MNT_ASYNC) | MNT_SOFTDEP; if ((mp->mnt_kern_flag & MNTK_SOFTDEP) == 0) { mp->mnt_kern_flag = (mp->mnt_kern_flag & ~MNTK_ASYNC) | MNTK_SOFTDEP | MNTK_NOASYNC; } ump = VFSTOUFS(mp); ump->um_softdep = sdp; MNT_IUNLOCK(mp); rw_init(LOCK_PTR(ump), "Per-Filesystem Softdep Lock"); sdp->sd_ump = ump; LIST_INIT(&ump->softdep_workitem_pending); LIST_INIT(&ump->softdep_journal_pending); TAILQ_INIT(&ump->softdep_unlinked); LIST_INIT(&ump->softdep_dirtycg); ump->softdep_worklist_tail = NULL; ump->softdep_on_worklist = 0; ump->softdep_deps = 0; LIST_INIT(&ump->softdep_mkdirlisthd); ump->pagedep_hashtbl = hashinit(desiredvnodes / 5, M_PAGEDEP, &ump->pagedep_hash_size); ump->pagedep_nextclean = 0; ump->inodedep_hashtbl = hashinit(desiredvnodes, M_INODEDEP, &ump->inodedep_hash_size); ump->inodedep_nextclean = 0; ump->newblk_hashtbl = hashinit(max_softdeps / 2, M_NEWBLK, &ump->newblk_hash_size); ump->bmsafemap_hashtbl = hashinit(1024, M_BMSAFEMAP, &ump->bmsafemap_hash_size); i = 1 << (ffs(desiredvnodes / 10) - 1); ump->indir_hashtbl = malloc(i * sizeof(struct indir_hashhead), M_FREEWORK, M_WAITOK); ump->indir_hash_size = i - 1; for (i = 0; i <= ump->indir_hash_size; i++) TAILQ_INIT(&ump->indir_hashtbl[i]); ACQUIRE_GBLLOCK(&lk); TAILQ_INSERT_TAIL(&softdepmounts, sdp, sd_next); FREE_GBLLOCK(&lk); if ((fs->fs_flags & FS_SUJ) && (error = journal_mount(mp, fs, cred)) != 0) { printf("Failed to start journal: %d\n", error); softdep_unmount(mp); return (error); } /* * Start our flushing thread in the bufdaemon process. */ ACQUIRE_LOCK(ump); ump->softdep_flags |= FLUSH_STARTING; FREE_LOCK(ump); kproc_kthread_add(&softdep_flush, mp, &bufdaemonproc, &ump->softdep_flushtd, 0, 0, "softdepflush", "%s worker", mp->mnt_stat.f_mntonname); ACQUIRE_LOCK(ump); while ((ump->softdep_flags & FLUSH_STARTING) != 0) { msleep(&ump->softdep_flushtd, LOCK_PTR(ump), PVM, "sdstart", hz / 2); } FREE_LOCK(ump); /* * When doing soft updates, the counters in the * superblock may have gotten out of sync. Recomputation * can take a long time and can be deferred for background * fsck. However, the old behavior of scanning the cylinder * groups and recalculating them at mount time is available * by setting vfs.ffs.compute_summary_at_mount to one. */ if (compute_summary_at_mount == 0 || fs->fs_clean != 0) return (0); bzero(&cstotal, sizeof cstotal); for (cyl = 0; cyl < fs->fs_ncg; cyl++) { if ((error = bread(devvp, fsbtodb(fs, cgtod(fs, cyl)), fs->fs_cgsize, cred, &bp)) != 0) { brelse(bp); softdep_unmount(mp); return (error); } cgp = (struct cg *)bp->b_data; cstotal.cs_nffree += cgp->cg_cs.cs_nffree; cstotal.cs_nbfree += cgp->cg_cs.cs_nbfree; cstotal.cs_nifree += cgp->cg_cs.cs_nifree; cstotal.cs_ndir += cgp->cg_cs.cs_ndir; fs->fs_cs(fs, cyl) = cgp->cg_cs; brelse(bp); } #ifdef DEBUG if (bcmp(&cstotal, &fs->fs_cstotal, sizeof cstotal)) printf("%s: superblock summary recomputed\n", fs->fs_fsmnt); #endif bcopy(&cstotal, &fs->fs_cstotal, sizeof cstotal); return (0); } void softdep_unmount(mp) struct mount *mp; { struct ufsmount *ump; #ifdef INVARIANTS int i; #endif KASSERT(MOUNTEDSOFTDEP(mp) != 0, ("softdep_unmount called on non-softdep filesystem")); ump = VFSTOUFS(mp); MNT_ILOCK(mp); mp->mnt_flag &= ~MNT_SOFTDEP; if (MOUNTEDSUJ(mp) == 0) { MNT_IUNLOCK(mp); } else { mp->mnt_flag &= ~MNT_SUJ; MNT_IUNLOCK(mp); journal_unmount(ump); } /* * Shut down our flushing thread. Check for NULL is if * softdep_mount errors out before the thread has been created. */ if (ump->softdep_flushtd != NULL) { ACQUIRE_LOCK(ump); ump->softdep_flags |= FLUSH_EXIT; wakeup(&ump->softdep_flushtd); msleep(&ump->softdep_flags, LOCK_PTR(ump), PVM | PDROP, "sdwait", 0); KASSERT((ump->softdep_flags & FLUSH_EXIT) == 0, ("Thread shutdown failed")); } /* * Free up our resources. */ ACQUIRE_GBLLOCK(&lk); TAILQ_REMOVE(&softdepmounts, ump->um_softdep, sd_next); FREE_GBLLOCK(&lk); rw_destroy(LOCK_PTR(ump)); hashdestroy(ump->pagedep_hashtbl, M_PAGEDEP, ump->pagedep_hash_size); hashdestroy(ump->inodedep_hashtbl, M_INODEDEP, ump->inodedep_hash_size); hashdestroy(ump->newblk_hashtbl, M_NEWBLK, ump->newblk_hash_size); hashdestroy(ump->bmsafemap_hashtbl, M_BMSAFEMAP, ump->bmsafemap_hash_size); free(ump->indir_hashtbl, M_FREEWORK); #ifdef INVARIANTS for (i = 0; i <= D_LAST; i++) KASSERT(ump->softdep_curdeps[i] == 0, ("Unmount %s: Dep type %s != 0 (%ld)", ump->um_fs->fs_fsmnt, TYPENAME(i), ump->softdep_curdeps[i])); #endif free(ump->um_softdep, M_MOUNTDATA); } static struct jblocks * jblocks_create(void) { struct jblocks *jblocks; jblocks = malloc(sizeof(*jblocks), M_JBLOCKS, M_WAITOK | M_ZERO); TAILQ_INIT(&jblocks->jb_segs); jblocks->jb_avail = 10; jblocks->jb_extent = malloc(sizeof(struct jextent) * jblocks->jb_avail, M_JBLOCKS, M_WAITOK | M_ZERO); return (jblocks); } static ufs2_daddr_t jblocks_alloc(jblocks, bytes, actual) struct jblocks *jblocks; int bytes; int *actual; { ufs2_daddr_t daddr; struct jextent *jext; int freecnt; int blocks; blocks = bytes / DEV_BSIZE; jext = &jblocks->jb_extent[jblocks->jb_head]; freecnt = jext->je_blocks - jblocks->jb_off; if (freecnt == 0) { jblocks->jb_off = 0; if (++jblocks->jb_head > jblocks->jb_used) jblocks->jb_head = 0; jext = &jblocks->jb_extent[jblocks->jb_head]; freecnt = jext->je_blocks; } if (freecnt > blocks) freecnt = blocks; *actual = freecnt * DEV_BSIZE; daddr = jext->je_daddr + jblocks->jb_off; jblocks->jb_off += freecnt; jblocks->jb_free -= freecnt; return (daddr); } static void jblocks_free(jblocks, mp, bytes) struct jblocks *jblocks; struct mount *mp; int bytes; { LOCK_OWNED(VFSTOUFS(mp)); jblocks->jb_free += bytes / DEV_BSIZE; if (jblocks->jb_suspended) worklist_speedup(mp); wakeup(jblocks); } static void jblocks_destroy(jblocks) struct jblocks *jblocks; { if (jblocks->jb_extent) free(jblocks->jb_extent, M_JBLOCKS); free(jblocks, M_JBLOCKS); } static void jblocks_add(jblocks, daddr, blocks) struct jblocks *jblocks; ufs2_daddr_t daddr; int blocks; { struct jextent *jext; jblocks->jb_blocks += blocks; jblocks->jb_free += blocks; jext = &jblocks->jb_extent[jblocks->jb_used]; /* Adding the first block. */ if (jext->je_daddr == 0) { jext->je_daddr = daddr; jext->je_blocks = blocks; return; } /* Extending the last extent. */ if (jext->je_daddr + jext->je_blocks == daddr) { jext->je_blocks += blocks; return; } /* Adding a new extent. */ if (++jblocks->jb_used == jblocks->jb_avail) { jblocks->jb_avail *= 2; jext = malloc(sizeof(struct jextent) * jblocks->jb_avail, M_JBLOCKS, M_WAITOK | M_ZERO); memcpy(jext, jblocks->jb_extent, sizeof(struct jextent) * jblocks->jb_used); free(jblocks->jb_extent, M_JBLOCKS); jblocks->jb_extent = jext; } jext = &jblocks->jb_extent[jblocks->jb_used]; jext->je_daddr = daddr; jext->je_blocks = blocks; return; } int softdep_journal_lookup(mp, vpp) struct mount *mp; struct vnode **vpp; { struct componentname cnp; struct vnode *dvp; ino_t sujournal; int error; error = VFS_VGET(mp, ROOTINO, LK_EXCLUSIVE, &dvp); if (error) return (error); bzero(&cnp, sizeof(cnp)); cnp.cn_nameiop = LOOKUP; cnp.cn_flags = ISLASTCN; cnp.cn_thread = curthread; cnp.cn_cred = curthread->td_ucred; cnp.cn_pnbuf = SUJ_FILE; cnp.cn_nameptr = SUJ_FILE; cnp.cn_namelen = strlen(SUJ_FILE); error = ufs_lookup_ino(dvp, NULL, &cnp, &sujournal); vput(dvp); if (error != 0) return (error); error = VFS_VGET(mp, sujournal, LK_EXCLUSIVE, vpp); return (error); } /* * Open and verify the journal file. */ static int journal_mount(mp, fs, cred) struct mount *mp; struct fs *fs; struct ucred *cred; { struct jblocks *jblocks; struct ufsmount *ump; struct vnode *vp; struct inode *ip; ufs2_daddr_t blkno; int bcount; int error; int i; ump = VFSTOUFS(mp); ump->softdep_journal_tail = NULL; ump->softdep_on_journal = 0; ump->softdep_accdeps = 0; ump->softdep_req = 0; ump->softdep_jblocks = NULL; error = softdep_journal_lookup(mp, &vp); if (error != 0) { printf("Failed to find journal. Use tunefs to create one\n"); return (error); } ip = VTOI(vp); if (ip->i_size < SUJ_MIN) { error = ENOSPC; goto out; } bcount = lblkno(fs, ip->i_size); /* Only use whole blocks. */ jblocks = jblocks_create(); for (i = 0; i < bcount; i++) { error = ufs_bmaparray(vp, i, &blkno, NULL, NULL, NULL); if (error) break; jblocks_add(jblocks, blkno, fsbtodb(fs, fs->fs_frag)); } if (error) { jblocks_destroy(jblocks); goto out; } jblocks->jb_low = jblocks->jb_free / 3; /* Reserve 33%. */ jblocks->jb_min = jblocks->jb_free / 10; /* Suspend at 10%. */ ump->softdep_jblocks = jblocks; out: if (error == 0) { MNT_ILOCK(mp); mp->mnt_flag |= MNT_SUJ; mp->mnt_flag &= ~MNT_SOFTDEP; MNT_IUNLOCK(mp); /* * Only validate the journal contents if the * filesystem is clean, otherwise we write the logs * but they'll never be used. If the filesystem was * still dirty when we mounted it the journal is * invalid and a new journal can only be valid if it * starts from a clean mount. */ if (fs->fs_clean) { DIP_SET(ip, i_modrev, fs->fs_mtime); ip->i_flags |= IN_MODIFIED; ffs_update(vp, 1); } } vput(vp); return (error); } static void journal_unmount(ump) struct ufsmount *ump; { if (ump->softdep_jblocks) jblocks_destroy(ump->softdep_jblocks); ump->softdep_jblocks = NULL; } /* * Called when a journal record is ready to be written. Space is allocated * and the journal entry is created when the journal is flushed to stable * store. */ static void add_to_journal(wk) struct worklist *wk; { struct ufsmount *ump; ump = VFSTOUFS(wk->wk_mp); LOCK_OWNED(ump); if (wk->wk_state & ONWORKLIST) panic("add_to_journal: %s(0x%X) already on list", TYPENAME(wk->wk_type), wk->wk_state); wk->wk_state |= ONWORKLIST | DEPCOMPLETE; if (LIST_EMPTY(&ump->softdep_journal_pending)) { ump->softdep_jblocks->jb_age = ticks; LIST_INSERT_HEAD(&ump->softdep_journal_pending, wk, wk_list); } else LIST_INSERT_AFTER(ump->softdep_journal_tail, wk, wk_list); ump->softdep_journal_tail = wk; ump->softdep_on_journal += 1; } /* * Remove an arbitrary item for the journal worklist maintain the tail * pointer. This happens when a new operation obviates the need to * journal an old operation. */ static void remove_from_journal(wk) struct worklist *wk; { struct ufsmount *ump; ump = VFSTOUFS(wk->wk_mp); LOCK_OWNED(ump); #ifdef SUJ_DEBUG { struct worklist *wkn; LIST_FOREACH(wkn, &ump->softdep_journal_pending, wk_list) if (wkn == wk) break; if (wkn == NULL) panic("remove_from_journal: %p is not in journal", wk); } #endif /* * We emulate a TAILQ to save space in most structures which do not * require TAILQ semantics. Here we must update the tail position * when removing the tail which is not the final entry. This works * only if the worklist linkage are at the beginning of the structure. */ if (ump->softdep_journal_tail == wk) ump->softdep_journal_tail = (struct worklist *)wk->wk_list.le_prev; WORKLIST_REMOVE(wk); ump->softdep_on_journal -= 1; } /* * Check for journal space as well as dependency limits so the prelink * code can throttle both journaled and non-journaled filesystems. * Threshold is 0 for low and 1 for min. */ static int journal_space(ump, thresh) struct ufsmount *ump; int thresh; { struct jblocks *jblocks; int limit, avail; jblocks = ump->softdep_jblocks; if (jblocks == NULL) return (1); /* * We use a tighter restriction here to prevent request_cleanup() * running in threads from running into locks we currently hold. * We have to be over the limit and our filesystem has to be * responsible for more than our share of that usage. */ limit = (max_softdeps / 10) * 9; if (dep_current[D_INODEDEP] > limit && ump->softdep_curdeps[D_INODEDEP] > limit / stat_flush_threads) return (0); if (thresh) thresh = jblocks->jb_min; else thresh = jblocks->jb_low; avail = (ump->softdep_on_journal * JREC_SIZE) / DEV_BSIZE; avail = jblocks->jb_free - avail; return (avail > thresh); } static void journal_suspend(ump) struct ufsmount *ump; { struct jblocks *jblocks; struct mount *mp; mp = UFSTOVFS(ump); jblocks = ump->softdep_jblocks; MNT_ILOCK(mp); if ((mp->mnt_kern_flag & MNTK_SUSPEND) == 0) { stat_journal_min++; mp->mnt_kern_flag |= MNTK_SUSPEND; mp->mnt_susp_owner = ump->softdep_flushtd; } jblocks->jb_suspended = 1; MNT_IUNLOCK(mp); } static int journal_unsuspend(struct ufsmount *ump) { struct jblocks *jblocks; struct mount *mp; mp = UFSTOVFS(ump); jblocks = ump->softdep_jblocks; if (jblocks != NULL && jblocks->jb_suspended && journal_space(ump, jblocks->jb_min)) { jblocks->jb_suspended = 0; FREE_LOCK(ump); mp->mnt_susp_owner = curthread; vfs_write_resume(mp, 0); ACQUIRE_LOCK(ump); return (1); } return (0); } /* * Called before any allocation function to be certain that there is * sufficient space in the journal prior to creating any new records. * Since in the case of block allocation we may have multiple locked * buffers at the time of the actual allocation we can not block * when the journal records are created. Doing so would create a deadlock * if any of these buffers needed to be flushed to reclaim space. Instead * we require a sufficiently large amount of available space such that * each thread in the system could have passed this allocation check and * still have sufficient free space. With 20% of a minimum journal size * of 1MB we have 6553 records available. */ int softdep_prealloc(vp, waitok) struct vnode *vp; int waitok; { struct ufsmount *ump; KASSERT(MOUNTEDSOFTDEP(vp->v_mount) != 0, ("softdep_prealloc called on non-softdep filesystem")); /* * Nothing to do if we are not running journaled soft updates. * If we currently hold the snapshot lock, we must avoid handling * other resources that could cause deadlock. */ if (DOINGSUJ(vp) == 0 || IS_SNAPSHOT(VTOI(vp))) return (0); ump = VFSTOUFS(vp->v_mount); ACQUIRE_LOCK(ump); if (journal_space(ump, 0)) { FREE_LOCK(ump); return (0); } stat_journal_low++; FREE_LOCK(ump); if (waitok == MNT_NOWAIT) return (ENOSPC); /* * Attempt to sync this vnode once to flush any journal * work attached to it. */ if ((curthread->td_pflags & TDP_COWINPROGRESS) == 0) ffs_syncvnode(vp, waitok, 0); ACQUIRE_LOCK(ump); process_removes(vp); process_truncates(vp); if (journal_space(ump, 0) == 0) { softdep_speedup(ump); if (journal_space(ump, 1) == 0) journal_suspend(ump); } FREE_LOCK(ump); return (0); } /* * Before adjusting a link count on a vnode verify that we have sufficient * journal space. If not, process operations that depend on the currently * locked pair of vnodes to try to flush space as the syncer, buf daemon, * and softdep flush threads can not acquire these locks to reclaim space. */ static void softdep_prelink(dvp, vp) struct vnode *dvp; struct vnode *vp; { struct ufsmount *ump; ump = VFSTOUFS(dvp->v_mount); LOCK_OWNED(ump); /* * Nothing to do if we have sufficient journal space. * If we currently hold the snapshot lock, we must avoid * handling other resources that could cause deadlock. */ if (journal_space(ump, 0) || (vp && IS_SNAPSHOT(VTOI(vp)))) return; stat_journal_low++; FREE_LOCK(ump); if (vp) ffs_syncvnode(vp, MNT_NOWAIT, 0); ffs_syncvnode(dvp, MNT_WAIT, 0); ACQUIRE_LOCK(ump); /* Process vp before dvp as it may create .. removes. */ if (vp) { process_removes(vp); process_truncates(vp); } process_removes(dvp); process_truncates(dvp); softdep_speedup(ump); process_worklist_item(UFSTOVFS(ump), 2, LK_NOWAIT); if (journal_space(ump, 0) == 0) { softdep_speedup(ump); if (journal_space(ump, 1) == 0) journal_suspend(ump); } } static void jseg_write(ump, jseg, data) struct ufsmount *ump; struct jseg *jseg; uint8_t *data; { struct jsegrec *rec; rec = (struct jsegrec *)data; rec->jsr_seq = jseg->js_seq; rec->jsr_oldest = jseg->js_oldseq; rec->jsr_cnt = jseg->js_cnt; rec->jsr_blocks = jseg->js_size / ump->um_devvp->v_bufobj.bo_bsize; rec->jsr_crc = 0; rec->jsr_time = ump->um_fs->fs_mtime; } static inline void inoref_write(inoref, jseg, rec) struct inoref *inoref; struct jseg *jseg; struct jrefrec *rec; { inoref->if_jsegdep->jd_seg = jseg; rec->jr_ino = inoref->if_ino; rec->jr_parent = inoref->if_parent; rec->jr_nlink = inoref->if_nlink; rec->jr_mode = inoref->if_mode; rec->jr_diroff = inoref->if_diroff; } static void jaddref_write(jaddref, jseg, data) struct jaddref *jaddref; struct jseg *jseg; uint8_t *data; { struct jrefrec *rec; rec = (struct jrefrec *)data; rec->jr_op = JOP_ADDREF; inoref_write(&jaddref->ja_ref, jseg, rec); } static void jremref_write(jremref, jseg, data) struct jremref *jremref; struct jseg *jseg; uint8_t *data; { struct jrefrec *rec; rec = (struct jrefrec *)data; rec->jr_op = JOP_REMREF; inoref_write(&jremref->jr_ref, jseg, rec); } static void jmvref_write(jmvref, jseg, data) struct jmvref *jmvref; struct jseg *jseg; uint8_t *data; { struct jmvrec *rec; rec = (struct jmvrec *)data; rec->jm_op = JOP_MVREF; rec->jm_ino = jmvref->jm_ino; rec->jm_parent = jmvref->jm_parent; rec->jm_oldoff = jmvref->jm_oldoff; rec->jm_newoff = jmvref->jm_newoff; } static void jnewblk_write(jnewblk, jseg, data) struct jnewblk *jnewblk; struct jseg *jseg; uint8_t *data; { struct jblkrec *rec; jnewblk->jn_jsegdep->jd_seg = jseg; rec = (struct jblkrec *)data; rec->jb_op = JOP_NEWBLK; rec->jb_ino = jnewblk->jn_ino; rec->jb_blkno = jnewblk->jn_blkno; rec->jb_lbn = jnewblk->jn_lbn; rec->jb_frags = jnewblk->jn_frags; rec->jb_oldfrags = jnewblk->jn_oldfrags; } static void jfreeblk_write(jfreeblk, jseg, data) struct jfreeblk *jfreeblk; struct jseg *jseg; uint8_t *data; { struct jblkrec *rec; jfreeblk->jf_dep.jb_jsegdep->jd_seg = jseg; rec = (struct jblkrec *)data; rec->jb_op = JOP_FREEBLK; rec->jb_ino = jfreeblk->jf_ino; rec->jb_blkno = jfreeblk->jf_blkno; rec->jb_lbn = jfreeblk->jf_lbn; rec->jb_frags = jfreeblk->jf_frags; rec->jb_oldfrags = 0; } static void jfreefrag_write(jfreefrag, jseg, data) struct jfreefrag *jfreefrag; struct jseg *jseg; uint8_t *data; { struct jblkrec *rec; jfreefrag->fr_jsegdep->jd_seg = jseg; rec = (struct jblkrec *)data; rec->jb_op = JOP_FREEBLK; rec->jb_ino = jfreefrag->fr_ino; rec->jb_blkno = jfreefrag->fr_blkno; rec->jb_lbn = jfreefrag->fr_lbn; rec->jb_frags = jfreefrag->fr_frags; rec->jb_oldfrags = 0; } static void jtrunc_write(jtrunc, jseg, data) struct jtrunc *jtrunc; struct jseg *jseg; uint8_t *data; { struct jtrncrec *rec; jtrunc->jt_dep.jb_jsegdep->jd_seg = jseg; rec = (struct jtrncrec *)data; rec->jt_op = JOP_TRUNC; rec->jt_ino = jtrunc->jt_ino; rec->jt_size = jtrunc->jt_size; rec->jt_extsize = jtrunc->jt_extsize; } static void jfsync_write(jfsync, jseg, data) struct jfsync *jfsync; struct jseg *jseg; uint8_t *data; { struct jtrncrec *rec; rec = (struct jtrncrec *)data; rec->jt_op = JOP_SYNC; rec->jt_ino = jfsync->jfs_ino; rec->jt_size = jfsync->jfs_size; rec->jt_extsize = jfsync->jfs_extsize; } static void softdep_flushjournal(mp) struct mount *mp; { struct jblocks *jblocks; struct ufsmount *ump; if (MOUNTEDSUJ(mp) == 0) return; ump = VFSTOUFS(mp); jblocks = ump->softdep_jblocks; ACQUIRE_LOCK(ump); while (ump->softdep_on_journal) { jblocks->jb_needseg = 1; softdep_process_journal(mp, NULL, MNT_WAIT); } FREE_LOCK(ump); } static void softdep_synchronize_completed(struct bio *); static void softdep_synchronize(struct bio *, struct ufsmount *, void *); static void softdep_synchronize_completed(bp) struct bio *bp; { struct jseg *oldest; struct jseg *jseg; struct ufsmount *ump; /* * caller1 marks the last segment written before we issued the * synchronize cache. */ jseg = bp->bio_caller1; if (jseg == NULL) { g_destroy_bio(bp); return; } ump = VFSTOUFS(jseg->js_list.wk_mp); ACQUIRE_LOCK(ump); oldest = NULL; /* * Mark all the journal entries waiting on the synchronize cache * as completed so they may continue on. */ while (jseg != NULL && (jseg->js_state & COMPLETE) == 0) { jseg->js_state |= COMPLETE; oldest = jseg; jseg = TAILQ_PREV(jseg, jseglst, js_next); } /* * Restart deferred journal entry processing from the oldest * completed jseg. */ if (oldest) complete_jsegs(oldest); FREE_LOCK(ump); g_destroy_bio(bp); } /* * Send BIO_FLUSH/SYNCHRONIZE CACHE to the device to enforce write ordering * barriers. The journal must be written prior to any blocks that depend * on it and the journal can not be released until the blocks have be * written. This code handles both barriers simultaneously. */ static void softdep_synchronize(bp, ump, caller1) struct bio *bp; struct ufsmount *ump; void *caller1; { bp->bio_cmd = BIO_FLUSH; bp->bio_flags |= BIO_ORDERED; bp->bio_data = NULL; bp->bio_offset = ump->um_cp->provider->mediasize; bp->bio_length = 0; bp->bio_done = softdep_synchronize_completed; bp->bio_caller1 = caller1; g_io_request(bp, (struct g_consumer *)ump->um_devvp->v_bufobj.bo_private); } /* * Flush some journal records to disk. */ static void softdep_process_journal(mp, needwk, flags) struct mount *mp; struct worklist *needwk; int flags; { struct jblocks *jblocks; struct ufsmount *ump; struct worklist *wk; struct jseg *jseg; struct buf *bp; struct bio *bio; uint8_t *data; struct fs *fs; int shouldflush; int segwritten; int jrecmin; /* Minimum records per block. */ int jrecmax; /* Maximum records per block. */ int size; int cnt; int off; int devbsize; if (MOUNTEDSUJ(mp) == 0) return; shouldflush = softdep_flushcache; bio = NULL; jseg = NULL; ump = VFSTOUFS(mp); LOCK_OWNED(ump); fs = ump->um_fs; jblocks = ump->softdep_jblocks; devbsize = ump->um_devvp->v_bufobj.bo_bsize; /* * We write anywhere between a disk block and fs block. The upper * bound is picked to prevent buffer cache fragmentation and limit * processing time per I/O. */ jrecmin = (devbsize / JREC_SIZE) - 1; /* -1 for seg header */ jrecmax = (fs->fs_bsize / devbsize) * jrecmin; segwritten = 0; for (;;) { cnt = ump->softdep_on_journal; /* * Criteria for writing a segment: * 1) We have a full block. * 2) We're called from jwait() and haven't found the * journal item yet. * 3) Always write if needseg is set. * 4) If we are called from process_worklist and have * not yet written anything we write a partial block * to enforce a 1 second maximum latency on journal * entries. */ if (cnt < (jrecmax - 1) && needwk == NULL && jblocks->jb_needseg == 0 && (segwritten || cnt == 0)) break; cnt++; /* * Verify some free journal space. softdep_prealloc() should * guarantee that we don't run out so this is indicative of * a problem with the flow control. Try to recover * gracefully in any event. */ while (jblocks->jb_free == 0) { if (flags != MNT_WAIT) break; printf("softdep: Out of journal space!\n"); softdep_speedup(ump); msleep(jblocks, LOCK_PTR(ump), PRIBIO, "jblocks", hz); } FREE_LOCK(ump); jseg = malloc(sizeof(*jseg), M_JSEG, M_SOFTDEP_FLAGS); workitem_alloc(&jseg->js_list, D_JSEG, mp); LIST_INIT(&jseg->js_entries); LIST_INIT(&jseg->js_indirs); jseg->js_state = ATTACHED; if (shouldflush == 0) jseg->js_state |= COMPLETE; else if (bio == NULL) bio = g_alloc_bio(); jseg->js_jblocks = jblocks; bp = geteblk(fs->fs_bsize, 0); ACQUIRE_LOCK(ump); /* * If there was a race while we were allocating the block * and jseg the entry we care about was likely written. * We bail out in both the WAIT and NOWAIT case and assume * the caller will loop if the entry it cares about is * not written. */ cnt = ump->softdep_on_journal; if (cnt + jblocks->jb_needseg == 0 || jblocks->jb_free == 0) { bp->b_flags |= B_INVAL | B_NOCACHE; WORKITEM_FREE(jseg, D_JSEG); FREE_LOCK(ump); brelse(bp); ACQUIRE_LOCK(ump); break; } /* * Calculate the disk block size required for the available * records rounded to the min size. */ if (cnt == 0) size = devbsize; else if (cnt < jrecmax) size = howmany(cnt, jrecmin) * devbsize; else size = fs->fs_bsize; /* * Allocate a disk block for this journal data and account * for truncation of the requested size if enough contiguous * space was not available. */ bp->b_blkno = jblocks_alloc(jblocks, size, &size); bp->b_lblkno = bp->b_blkno; bp->b_offset = bp->b_blkno * DEV_BSIZE; bp->b_bcount = size; bp->b_flags &= ~B_INVAL; bp->b_flags |= B_VALIDSUSPWRT | B_NOCOPY; /* * Initialize our jseg with cnt records. Assign the next * sequence number to it and link it in-order. */ cnt = MIN(cnt, (size / devbsize) * jrecmin); jseg->js_buf = bp; jseg->js_cnt = cnt; jseg->js_refs = cnt + 1; /* Self ref. */ jseg->js_size = size; jseg->js_seq = jblocks->jb_nextseq++; if (jblocks->jb_oldestseg == NULL) jblocks->jb_oldestseg = jseg; jseg->js_oldseq = jblocks->jb_oldestseg->js_seq; TAILQ_INSERT_TAIL(&jblocks->jb_segs, jseg, js_next); if (jblocks->jb_writeseg == NULL) jblocks->jb_writeseg = jseg; /* * Start filling in records from the pending list. */ data = bp->b_data; off = 0; /* * Always put a header on the first block. * XXX As with below, there might not be a chance to get * into the loop. Ensure that something valid is written. */ jseg_write(ump, jseg, data); off += JREC_SIZE; data = bp->b_data + off; /* * XXX Something is wrong here. There's no work to do, * but we need to perform and I/O and allow it to complete * anyways. */ if (LIST_EMPTY(&ump->softdep_journal_pending)) stat_emptyjblocks++; while ((wk = LIST_FIRST(&ump->softdep_journal_pending)) != NULL) { if (cnt == 0) break; /* Place a segment header on every device block. */ if ((off % devbsize) == 0) { jseg_write(ump, jseg, data); off += JREC_SIZE; data = bp->b_data + off; } if (wk == needwk) needwk = NULL; remove_from_journal(wk); wk->wk_state |= INPROGRESS; WORKLIST_INSERT(&jseg->js_entries, wk); switch (wk->wk_type) { case D_JADDREF: jaddref_write(WK_JADDREF(wk), jseg, data); break; case D_JREMREF: jremref_write(WK_JREMREF(wk), jseg, data); break; case D_JMVREF: jmvref_write(WK_JMVREF(wk), jseg, data); break; case D_JNEWBLK: jnewblk_write(WK_JNEWBLK(wk), jseg, data); break; case D_JFREEBLK: jfreeblk_write(WK_JFREEBLK(wk), jseg, data); break; case D_JFREEFRAG: jfreefrag_write(WK_JFREEFRAG(wk), jseg, data); break; case D_JTRUNC: jtrunc_write(WK_JTRUNC(wk), jseg, data); break; case D_JFSYNC: jfsync_write(WK_JFSYNC(wk), jseg, data); break; default: panic("process_journal: Unknown type %s", TYPENAME(wk->wk_type)); /* NOTREACHED */ } off += JREC_SIZE; data = bp->b_data + off; cnt--; } /* Clear any remaining space so we don't leak kernel data */ if (size > off) bzero(data, size - off); /* * Write this one buffer and continue. */ segwritten = 1; jblocks->jb_needseg = 0; WORKLIST_INSERT(&bp->b_dep, &jseg->js_list); FREE_LOCK(ump); pbgetvp(ump->um_devvp, bp); /* * We only do the blocking wait once we find the journal * entry we're looking for. */ if (needwk == NULL && flags == MNT_WAIT) bwrite(bp); else bawrite(bp); ACQUIRE_LOCK(ump); } /* * If we wrote a segment issue a synchronize cache so the journal * is reflected on disk before the data is written. Since reclaiming * journal space also requires writing a journal record this * process also enforces a barrier before reclamation. */ if (segwritten && shouldflush) { softdep_synchronize(bio, ump, TAILQ_LAST(&jblocks->jb_segs, jseglst)); } else if (bio) g_destroy_bio(bio); /* * If we've suspended the filesystem because we ran out of journal * space either try to sync it here to make some progress or * unsuspend it if we already have. */ if (flags == 0 && jblocks->jb_suspended) { if (journal_unsuspend(ump)) return; FREE_LOCK(ump); VFS_SYNC(mp, MNT_NOWAIT); ffs_sbupdate(ump, MNT_WAIT, 0); ACQUIRE_LOCK(ump); } } /* * Complete a jseg, allowing all dependencies awaiting journal writes * to proceed. Each journal dependency also attaches a jsegdep to dependent * structures so that the journal segment can be freed to reclaim space. */ static void complete_jseg(jseg) struct jseg *jseg; { struct worklist *wk; struct jmvref *jmvref; int waiting; #ifdef INVARIANTS int i = 0; #endif while ((wk = LIST_FIRST(&jseg->js_entries)) != NULL) { WORKLIST_REMOVE(wk); waiting = wk->wk_state & IOWAITING; wk->wk_state &= ~(INPROGRESS | IOWAITING); wk->wk_state |= COMPLETE; KASSERT(i++ < jseg->js_cnt, ("handle_written_jseg: overflow %d >= %d", i - 1, jseg->js_cnt)); switch (wk->wk_type) { case D_JADDREF: handle_written_jaddref(WK_JADDREF(wk)); break; case D_JREMREF: handle_written_jremref(WK_JREMREF(wk)); break; case D_JMVREF: rele_jseg(jseg); /* No jsegdep. */ jmvref = WK_JMVREF(wk); LIST_REMOVE(jmvref, jm_deps); if ((jmvref->jm_pagedep->pd_state & ONWORKLIST) == 0) free_pagedep(jmvref->jm_pagedep); WORKITEM_FREE(jmvref, D_JMVREF); break; case D_JNEWBLK: handle_written_jnewblk(WK_JNEWBLK(wk)); break; case D_JFREEBLK: handle_written_jblkdep(&WK_JFREEBLK(wk)->jf_dep); break; case D_JTRUNC: handle_written_jblkdep(&WK_JTRUNC(wk)->jt_dep); break; case D_JFSYNC: rele_jseg(jseg); /* No jsegdep. */ WORKITEM_FREE(wk, D_JFSYNC); break; case D_JFREEFRAG: handle_written_jfreefrag(WK_JFREEFRAG(wk)); break; default: panic("handle_written_jseg: Unknown type %s", TYPENAME(wk->wk_type)); /* NOTREACHED */ } if (waiting) wakeup(wk); } /* Release the self reference so the structure may be freed. */ rele_jseg(jseg); } /* * Determine which jsegs are ready for completion processing. Waits for * synchronize cache to complete as well as forcing in-order completion * of journal entries. */ static void complete_jsegs(jseg) struct jseg *jseg; { struct jblocks *jblocks; struct jseg *jsegn; jblocks = jseg->js_jblocks; /* * Don't allow out of order completions. If this isn't the first * block wait for it to write before we're done. */ if (jseg != jblocks->jb_writeseg) return; /* Iterate through available jsegs processing their entries. */ while (jseg && (jseg->js_state & ALLCOMPLETE) == ALLCOMPLETE) { jblocks->jb_oldestwrseq = jseg->js_oldseq; jsegn = TAILQ_NEXT(jseg, js_next); complete_jseg(jseg); jseg = jsegn; } jblocks->jb_writeseg = jseg; /* * Attempt to free jsegs now that oldestwrseq may have advanced. */ free_jsegs(jblocks); } /* * Mark a jseg as DEPCOMPLETE and throw away the buffer. Attempt to handle * the final completions. */ static void handle_written_jseg(jseg, bp) struct jseg *jseg; struct buf *bp; { if (jseg->js_refs == 0) panic("handle_written_jseg: No self-reference on %p", jseg); jseg->js_state |= DEPCOMPLETE; /* * We'll never need this buffer again, set flags so it will be * discarded. */ bp->b_flags |= B_INVAL | B_NOCACHE; pbrelvp(bp); complete_jsegs(jseg); } static inline struct jsegdep * inoref_jseg(inoref) struct inoref *inoref; { struct jsegdep *jsegdep; jsegdep = inoref->if_jsegdep; inoref->if_jsegdep = NULL; return (jsegdep); } /* * Called once a jremref has made it to stable store. The jremref is marked * complete and we attempt to free it. Any pagedeps writes sleeping waiting * for the jremref to complete will be awoken by free_jremref. */ static void handle_written_jremref(jremref) struct jremref *jremref; { struct inodedep *inodedep; struct jsegdep *jsegdep; struct dirrem *dirrem; /* Grab the jsegdep. */ jsegdep = inoref_jseg(&jremref->jr_ref); /* * Remove us from the inoref list. */ if (inodedep_lookup(jremref->jr_list.wk_mp, jremref->jr_ref.if_ino, 0, &inodedep) == 0) panic("handle_written_jremref: Lost inodedep"); TAILQ_REMOVE(&inodedep->id_inoreflst, &jremref->jr_ref, if_deps); /* * Complete the dirrem. */ dirrem = jremref->jr_dirrem; jremref->jr_dirrem = NULL; LIST_REMOVE(jremref, jr_deps); jsegdep->jd_state |= jremref->jr_state & MKDIR_PARENT; jwork_insert(&dirrem->dm_jwork, jsegdep); if (LIST_EMPTY(&dirrem->dm_jremrefhd) && (dirrem->dm_state & COMPLETE) != 0) add_to_worklist(&dirrem->dm_list, 0); free_jremref(jremref); } /* * Called once a jaddref has made it to stable store. The dependency is * marked complete and any dependent structures are added to the inode * bufwait list to be completed as soon as it is written. If a bitmap write * depends on this entry we move the inode into the inodedephd of the * bmsafemap dependency and attempt to remove the jaddref from the bmsafemap. */ static void handle_written_jaddref(jaddref) struct jaddref *jaddref; { struct jsegdep *jsegdep; struct inodedep *inodedep; struct diradd *diradd; struct mkdir *mkdir; /* Grab the jsegdep. */ jsegdep = inoref_jseg(&jaddref->ja_ref); mkdir = NULL; diradd = NULL; if (inodedep_lookup(jaddref->ja_list.wk_mp, jaddref->ja_ino, 0, &inodedep) == 0) panic("handle_written_jaddref: Lost inodedep."); if (jaddref->ja_diradd == NULL) panic("handle_written_jaddref: No dependency"); if (jaddref->ja_diradd->da_list.wk_type == D_DIRADD) { diradd = jaddref->ja_diradd; WORKLIST_INSERT(&inodedep->id_bufwait, &diradd->da_list); } else if (jaddref->ja_state & MKDIR_PARENT) { mkdir = jaddref->ja_mkdir; WORKLIST_INSERT(&inodedep->id_bufwait, &mkdir->md_list); } else if (jaddref->ja_state & MKDIR_BODY) mkdir = jaddref->ja_mkdir; else panic("handle_written_jaddref: Unknown dependency %p", jaddref->ja_diradd); jaddref->ja_diradd = NULL; /* also clears ja_mkdir */ /* * Remove us from the inode list. */ TAILQ_REMOVE(&inodedep->id_inoreflst, &jaddref->ja_ref, if_deps); /* * The mkdir may be waiting on the jaddref to clear before freeing. */ if (mkdir) { KASSERT(mkdir->md_list.wk_type == D_MKDIR, ("handle_written_jaddref: Incorrect type for mkdir %s", TYPENAME(mkdir->md_list.wk_type))); mkdir->md_jaddref = NULL; diradd = mkdir->md_diradd; mkdir->md_state |= DEPCOMPLETE; complete_mkdir(mkdir); } jwork_insert(&diradd->da_jwork, jsegdep); if (jaddref->ja_state & NEWBLOCK) { inodedep->id_state |= ONDEPLIST; LIST_INSERT_HEAD(&inodedep->id_bmsafemap->sm_inodedephd, inodedep, id_deps); } free_jaddref(jaddref); } /* * Called once a jnewblk journal is written. The allocdirect or allocindir * is placed in the bmsafemap to await notification of a written bitmap. If * the operation was canceled we add the segdep to the appropriate * dependency to free the journal space once the canceling operation * completes. */ static void handle_written_jnewblk(jnewblk) struct jnewblk *jnewblk; { struct bmsafemap *bmsafemap; struct freefrag *freefrag; struct freework *freework; struct jsegdep *jsegdep; struct newblk *newblk; /* Grab the jsegdep. */ jsegdep = jnewblk->jn_jsegdep; jnewblk->jn_jsegdep = NULL; if (jnewblk->jn_dep == NULL) panic("handle_written_jnewblk: No dependency for the segdep."); switch (jnewblk->jn_dep->wk_type) { case D_NEWBLK: case D_ALLOCDIRECT: case D_ALLOCINDIR: /* * Add the written block to the bmsafemap so it can * be notified when the bitmap is on disk. */ newblk = WK_NEWBLK(jnewblk->jn_dep); newblk->nb_jnewblk = NULL; if ((newblk->nb_state & GOINGAWAY) == 0) { bmsafemap = newblk->nb_bmsafemap; newblk->nb_state |= ONDEPLIST; LIST_INSERT_HEAD(&bmsafemap->sm_newblkhd, newblk, nb_deps); } jwork_insert(&newblk->nb_jwork, jsegdep); break; case D_FREEFRAG: /* * A newblock being removed by a freefrag when replaced by * frag extension. */ freefrag = WK_FREEFRAG(jnewblk->jn_dep); freefrag->ff_jdep = NULL; jwork_insert(&freefrag->ff_jwork, jsegdep); break; case D_FREEWORK: /* * A direct block was removed by truncate. */ freework = WK_FREEWORK(jnewblk->jn_dep); freework->fw_jnewblk = NULL; jwork_insert(&freework->fw_freeblks->fb_jwork, jsegdep); break; default: panic("handle_written_jnewblk: Unknown type %d.", jnewblk->jn_dep->wk_type); } jnewblk->jn_dep = NULL; free_jnewblk(jnewblk); } /* * Cancel a jfreefrag that won't be needed, probably due to colliding with * an in-flight allocation that has not yet been committed. Divorce us * from the freefrag and mark it DEPCOMPLETE so that it may be added * to the worklist. */ static void cancel_jfreefrag(jfreefrag) struct jfreefrag *jfreefrag; { struct freefrag *freefrag; if (jfreefrag->fr_jsegdep) { free_jsegdep(jfreefrag->fr_jsegdep); jfreefrag->fr_jsegdep = NULL; } freefrag = jfreefrag->fr_freefrag; jfreefrag->fr_freefrag = NULL; free_jfreefrag(jfreefrag); freefrag->ff_state |= DEPCOMPLETE; CTR1(KTR_SUJ, "cancel_jfreefrag: blkno %jd", freefrag->ff_blkno); } /* * Free a jfreefrag when the parent freefrag is rendered obsolete. */ static void free_jfreefrag(jfreefrag) struct jfreefrag *jfreefrag; { if (jfreefrag->fr_state & INPROGRESS) WORKLIST_REMOVE(&jfreefrag->fr_list); else if (jfreefrag->fr_state & ONWORKLIST) remove_from_journal(&jfreefrag->fr_list); if (jfreefrag->fr_freefrag != NULL) panic("free_jfreefrag: Still attached to a freefrag."); WORKITEM_FREE(jfreefrag, D_JFREEFRAG); } /* * Called when the journal write for a jfreefrag completes. The parent * freefrag is added to the worklist if this completes its dependencies. */ static void handle_written_jfreefrag(jfreefrag) struct jfreefrag *jfreefrag; { struct jsegdep *jsegdep; struct freefrag *freefrag; /* Grab the jsegdep. */ jsegdep = jfreefrag->fr_jsegdep; jfreefrag->fr_jsegdep = NULL; freefrag = jfreefrag->fr_freefrag; if (freefrag == NULL) panic("handle_written_jfreefrag: No freefrag."); freefrag->ff_state |= DEPCOMPLETE; freefrag->ff_jdep = NULL; jwork_insert(&freefrag->ff_jwork, jsegdep); if ((freefrag->ff_state & ALLCOMPLETE) == ALLCOMPLETE) add_to_worklist(&freefrag->ff_list, 0); jfreefrag->fr_freefrag = NULL; free_jfreefrag(jfreefrag); } /* * Called when the journal write for a jfreeblk completes. The jfreeblk * is removed from the freeblks list of pending journal writes and the * jsegdep is moved to the freeblks jwork to be completed when all blocks * have been reclaimed. */ static void handle_written_jblkdep(jblkdep) struct jblkdep *jblkdep; { struct freeblks *freeblks; struct jsegdep *jsegdep; /* Grab the jsegdep. */ jsegdep = jblkdep->jb_jsegdep; jblkdep->jb_jsegdep = NULL; freeblks = jblkdep->jb_freeblks; LIST_REMOVE(jblkdep, jb_deps); jwork_insert(&freeblks->fb_jwork, jsegdep); /* * If the freeblks is all journaled, we can add it to the worklist. */ if (LIST_EMPTY(&freeblks->fb_jblkdephd) && (freeblks->fb_state & ALLCOMPLETE) == ALLCOMPLETE) add_to_worklist(&freeblks->fb_list, WK_NODELAY); free_jblkdep(jblkdep); } static struct jsegdep * newjsegdep(struct worklist *wk) { struct jsegdep *jsegdep; jsegdep = malloc(sizeof(*jsegdep), M_JSEGDEP, M_SOFTDEP_FLAGS); workitem_alloc(&jsegdep->jd_list, D_JSEGDEP, wk->wk_mp); jsegdep->jd_seg = NULL; return (jsegdep); } static struct jmvref * newjmvref(dp, ino, oldoff, newoff) struct inode *dp; ino_t ino; off_t oldoff; off_t newoff; { struct jmvref *jmvref; jmvref = malloc(sizeof(*jmvref), M_JMVREF, M_SOFTDEP_FLAGS); workitem_alloc(&jmvref->jm_list, D_JMVREF, UFSTOVFS(dp->i_ump)); jmvref->jm_list.wk_state = ATTACHED | DEPCOMPLETE; jmvref->jm_parent = dp->i_number; jmvref->jm_ino = ino; jmvref->jm_oldoff = oldoff; jmvref->jm_newoff = newoff; return (jmvref); } /* * Allocate a new jremref that tracks the removal of ip from dp with the * directory entry offset of diroff. Mark the entry as ATTACHED and * DEPCOMPLETE as we have all the information required for the journal write * and the directory has already been removed from the buffer. The caller * is responsible for linking the jremref into the pagedep and adding it * to the journal to write. The MKDIR_PARENT flag is set if we're doing * a DOTDOT addition so handle_workitem_remove() can properly assign * the jsegdep when we're done. */ static struct jremref * newjremref(struct dirrem *dirrem, struct inode *dp, struct inode *ip, off_t diroff, nlink_t nlink) { struct jremref *jremref; jremref = malloc(sizeof(*jremref), M_JREMREF, M_SOFTDEP_FLAGS); workitem_alloc(&jremref->jr_list, D_JREMREF, UFSTOVFS(dp->i_ump)); jremref->jr_state = ATTACHED; newinoref(&jremref->jr_ref, ip->i_number, dp->i_number, diroff, nlink, ip->i_mode); jremref->jr_dirrem = dirrem; return (jremref); } static inline void newinoref(struct inoref *inoref, ino_t ino, ino_t parent, off_t diroff, nlink_t nlink, uint16_t mode) { inoref->if_jsegdep = newjsegdep(&inoref->if_list); inoref->if_diroff = diroff; inoref->if_ino = ino; inoref->if_parent = parent; inoref->if_nlink = nlink; inoref->if_mode = mode; } /* * Allocate a new jaddref to track the addition of ino to dp at diroff. The * directory offset may not be known until later. The caller is responsible * adding the entry to the journal when this information is available. nlink * should be the link count prior to the addition and mode is only required * to have the correct FMT. */ static struct jaddref * newjaddref(struct inode *dp, ino_t ino, off_t diroff, int16_t nlink, uint16_t mode) { struct jaddref *jaddref; jaddref = malloc(sizeof(*jaddref), M_JADDREF, M_SOFTDEP_FLAGS); workitem_alloc(&jaddref->ja_list, D_JADDREF, UFSTOVFS(dp->i_ump)); jaddref->ja_state = ATTACHED; jaddref->ja_mkdir = NULL; newinoref(&jaddref->ja_ref, ino, dp->i_number, diroff, nlink, mode); return (jaddref); } /* * Create a new free dependency for a freework. The caller is responsible * for adjusting the reference count when it has the lock held. The freedep * will track an outstanding bitmap write that will ultimately clear the * freework to continue. */ static struct freedep * newfreedep(struct freework *freework) { struct freedep *freedep; freedep = malloc(sizeof(*freedep), M_FREEDEP, M_SOFTDEP_FLAGS); workitem_alloc(&freedep->fd_list, D_FREEDEP, freework->fw_list.wk_mp); freedep->fd_freework = freework; return (freedep); } /* * Free a freedep structure once the buffer it is linked to is written. If * this is the last reference to the freework schedule it for completion. */ static void free_freedep(freedep) struct freedep *freedep; { struct freework *freework; freework = freedep->fd_freework; freework->fw_freeblks->fb_cgwait--; if (--freework->fw_ref == 0) freework_enqueue(freework); WORKITEM_FREE(freedep, D_FREEDEP); } /* * Allocate a new freework structure that may be a level in an indirect * when parent is not NULL or a top level block when it is. The top level * freework structures are allocated without the per-filesystem lock held * and before the freeblks is visible outside of softdep_setup_freeblocks(). */ static struct freework * newfreework(ump, freeblks, parent, lbn, nb, frags, off, journal) struct ufsmount *ump; struct freeblks *freeblks; struct freework *parent; ufs_lbn_t lbn; ufs2_daddr_t nb; int frags; int off; int journal; { struct freework *freework; freework = malloc(sizeof(*freework), M_FREEWORK, M_SOFTDEP_FLAGS); workitem_alloc(&freework->fw_list, D_FREEWORK, freeblks->fb_list.wk_mp); freework->fw_state = ATTACHED; freework->fw_jnewblk = NULL; freework->fw_freeblks = freeblks; freework->fw_parent = parent; freework->fw_lbn = lbn; freework->fw_blkno = nb; freework->fw_frags = frags; freework->fw_indir = NULL; freework->fw_ref = (MOUNTEDSUJ(UFSTOVFS(ump)) == 0 || lbn >= -NXADDR) ? 0 : NINDIR(ump->um_fs) + 1; freework->fw_start = freework->fw_off = off; if (journal) newjfreeblk(freeblks, lbn, nb, frags); if (parent == NULL) { ACQUIRE_LOCK(ump); WORKLIST_INSERT(&freeblks->fb_freeworkhd, &freework->fw_list); freeblks->fb_ref++; FREE_LOCK(ump); } return (freework); } /* * Eliminate a jfreeblk for a block that does not need journaling. */ static void cancel_jfreeblk(freeblks, blkno) struct freeblks *freeblks; ufs2_daddr_t blkno; { struct jfreeblk *jfreeblk; struct jblkdep *jblkdep; LIST_FOREACH(jblkdep, &freeblks->fb_jblkdephd, jb_deps) { if (jblkdep->jb_list.wk_type != D_JFREEBLK) continue; jfreeblk = WK_JFREEBLK(&jblkdep->jb_list); if (jfreeblk->jf_blkno == blkno) break; } if (jblkdep == NULL) return; CTR1(KTR_SUJ, "cancel_jfreeblk: blkno %jd", blkno); free_jsegdep(jblkdep->jb_jsegdep); LIST_REMOVE(jblkdep, jb_deps); WORKITEM_FREE(jfreeblk, D_JFREEBLK); } /* * Allocate a new jfreeblk to journal top level block pointer when truncating * a file. The caller must add this to the worklist when the per-filesystem * lock is held. */ static struct jfreeblk * newjfreeblk(freeblks, lbn, blkno, frags) struct freeblks *freeblks; ufs_lbn_t lbn; ufs2_daddr_t blkno; int frags; { struct jfreeblk *jfreeblk; jfreeblk = malloc(sizeof(*jfreeblk), M_JFREEBLK, M_SOFTDEP_FLAGS); workitem_alloc(&jfreeblk->jf_dep.jb_list, D_JFREEBLK, freeblks->fb_list.wk_mp); jfreeblk->jf_dep.jb_jsegdep = newjsegdep(&jfreeblk->jf_dep.jb_list); jfreeblk->jf_dep.jb_freeblks = freeblks; jfreeblk->jf_ino = freeblks->fb_inum; jfreeblk->jf_lbn = lbn; jfreeblk->jf_blkno = blkno; jfreeblk->jf_frags = frags; LIST_INSERT_HEAD(&freeblks->fb_jblkdephd, &jfreeblk->jf_dep, jb_deps); return (jfreeblk); } /* * The journal is only prepared to handle full-size block numbers, so we * have to adjust the record to reflect the change to a full-size block. * For example, suppose we have a block made up of fragments 8-15 and * want to free its last two fragments. We are given a request that says: * FREEBLK ino=5, blkno=14, lbn=0, frags=2, oldfrags=0 * where frags are the number of fragments to free and oldfrags are the * number of fragments to keep. To block align it, we have to change it to * have a valid full-size blkno, so it becomes: * FREEBLK ino=5, blkno=8, lbn=0, frags=2, oldfrags=6 */ static void adjust_newfreework(freeblks, frag_offset) struct freeblks *freeblks; int frag_offset; { struct jfreeblk *jfreeblk; KASSERT((LIST_FIRST(&freeblks->fb_jblkdephd) != NULL && LIST_FIRST(&freeblks->fb_jblkdephd)->jb_list.wk_type == D_JFREEBLK), ("adjust_newfreework: Missing freeblks dependency")); jfreeblk = WK_JFREEBLK(LIST_FIRST(&freeblks->fb_jblkdephd)); jfreeblk->jf_blkno -= frag_offset; jfreeblk->jf_frags += frag_offset; } /* * Allocate a new jtrunc to track a partial truncation. */ static struct jtrunc * newjtrunc(freeblks, size, extsize) struct freeblks *freeblks; off_t size; int extsize; { struct jtrunc *jtrunc; jtrunc = malloc(sizeof(*jtrunc), M_JTRUNC, M_SOFTDEP_FLAGS); workitem_alloc(&jtrunc->jt_dep.jb_list, D_JTRUNC, freeblks->fb_list.wk_mp); jtrunc->jt_dep.jb_jsegdep = newjsegdep(&jtrunc->jt_dep.jb_list); jtrunc->jt_dep.jb_freeblks = freeblks; jtrunc->jt_ino = freeblks->fb_inum; jtrunc->jt_size = size; jtrunc->jt_extsize = extsize; LIST_INSERT_HEAD(&freeblks->fb_jblkdephd, &jtrunc->jt_dep, jb_deps); return (jtrunc); } /* * If we're canceling a new bitmap we have to search for another ref * to move into the bmsafemap dep. This might be better expressed * with another structure. */ static void move_newblock_dep(jaddref, inodedep) struct jaddref *jaddref; struct inodedep *inodedep; { struct inoref *inoref; struct jaddref *jaddrefn; jaddrefn = NULL; for (inoref = TAILQ_NEXT(&jaddref->ja_ref, if_deps); inoref; inoref = TAILQ_NEXT(inoref, if_deps)) { if ((jaddref->ja_state & NEWBLOCK) && inoref->if_list.wk_type == D_JADDREF) { jaddrefn = (struct jaddref *)inoref; break; } } if (jaddrefn == NULL) return; jaddrefn->ja_state &= ~(ATTACHED | UNDONE); jaddrefn->ja_state |= jaddref->ja_state & (ATTACHED | UNDONE | NEWBLOCK); jaddref->ja_state &= ~(ATTACHED | UNDONE | NEWBLOCK); jaddref->ja_state |= ATTACHED; LIST_REMOVE(jaddref, ja_bmdeps); LIST_INSERT_HEAD(&inodedep->id_bmsafemap->sm_jaddrefhd, jaddrefn, ja_bmdeps); } /* * Cancel a jaddref either before it has been written or while it is being * written. This happens when a link is removed before the add reaches * the disk. The jaddref dependency is kept linked into the bmsafemap * and inode to prevent the link count or bitmap from reaching the disk * until handle_workitem_remove() re-adjusts the counts and bitmaps as * required. * * Returns 1 if the canceled addref requires journaling of the remove and * 0 otherwise. */ static int cancel_jaddref(jaddref, inodedep, wkhd) struct jaddref *jaddref; struct inodedep *inodedep; struct workhead *wkhd; { struct inoref *inoref; struct jsegdep *jsegdep; int needsj; KASSERT((jaddref->ja_state & COMPLETE) == 0, ("cancel_jaddref: Canceling complete jaddref")); if (jaddref->ja_state & (INPROGRESS | COMPLETE)) needsj = 1; else needsj = 0; if (inodedep == NULL) if (inodedep_lookup(jaddref->ja_list.wk_mp, jaddref->ja_ino, 0, &inodedep) == 0) panic("cancel_jaddref: Lost inodedep"); /* * We must adjust the nlink of any reference operation that follows * us so that it is consistent with the in-memory reference. This * ensures that inode nlink rollbacks always have the correct link. */ if (needsj == 0) { for (inoref = TAILQ_NEXT(&jaddref->ja_ref, if_deps); inoref; inoref = TAILQ_NEXT(inoref, if_deps)) { if (inoref->if_state & GOINGAWAY) break; inoref->if_nlink--; } } jsegdep = inoref_jseg(&jaddref->ja_ref); if (jaddref->ja_state & NEWBLOCK) move_newblock_dep(jaddref, inodedep); wake_worklist(&jaddref->ja_list); jaddref->ja_mkdir = NULL; if (jaddref->ja_state & INPROGRESS) { jaddref->ja_state &= ~INPROGRESS; WORKLIST_REMOVE(&jaddref->ja_list); jwork_insert(wkhd, jsegdep); } else { free_jsegdep(jsegdep); if (jaddref->ja_state & DEPCOMPLETE) remove_from_journal(&jaddref->ja_list); } jaddref->ja_state |= (GOINGAWAY | DEPCOMPLETE); /* * Leave NEWBLOCK jaddrefs on the inodedep so handle_workitem_remove * can arrange for them to be freed with the bitmap. Otherwise we * no longer need this addref attached to the inoreflst and it * will incorrectly adjust nlink if we leave it. */ if ((jaddref->ja_state & NEWBLOCK) == 0) { TAILQ_REMOVE(&inodedep->id_inoreflst, &jaddref->ja_ref, if_deps); jaddref->ja_state |= COMPLETE; free_jaddref(jaddref); return (needsj); } /* * Leave the head of the list for jsegdeps for fast merging. */ if (LIST_FIRST(wkhd) != NULL) { jaddref->ja_state |= ONWORKLIST; LIST_INSERT_AFTER(LIST_FIRST(wkhd), &jaddref->ja_list, wk_list); } else WORKLIST_INSERT(wkhd, &jaddref->ja_list); return (needsj); } /* * Attempt to free a jaddref structure when some work completes. This * should only succeed once the entry is written and all dependencies have * been notified. */ static void free_jaddref(jaddref) struct jaddref *jaddref; { if ((jaddref->ja_state & ALLCOMPLETE) != ALLCOMPLETE) return; if (jaddref->ja_ref.if_jsegdep) panic("free_jaddref: segdep attached to jaddref %p(0x%X)\n", jaddref, jaddref->ja_state); if (jaddref->ja_state & NEWBLOCK) LIST_REMOVE(jaddref, ja_bmdeps); if (jaddref->ja_state & (INPROGRESS | ONWORKLIST)) panic("free_jaddref: Bad state %p(0x%X)", jaddref, jaddref->ja_state); if (jaddref->ja_mkdir != NULL) panic("free_jaddref: Work pending, 0x%X\n", jaddref->ja_state); WORKITEM_FREE(jaddref, D_JADDREF); } /* * Free a jremref structure once it has been written or discarded. */ static void free_jremref(jremref) struct jremref *jremref; { if (jremref->jr_ref.if_jsegdep) free_jsegdep(jremref->jr_ref.if_jsegdep); if (jremref->jr_state & INPROGRESS) panic("free_jremref: IO still pending"); WORKITEM_FREE(jremref, D_JREMREF); } /* * Free a jnewblk structure. */ static void free_jnewblk(jnewblk) struct jnewblk *jnewblk; { if ((jnewblk->jn_state & ALLCOMPLETE) != ALLCOMPLETE) return; LIST_REMOVE(jnewblk, jn_deps); if (jnewblk->jn_dep != NULL) panic("free_jnewblk: Dependency still attached."); WORKITEM_FREE(jnewblk, D_JNEWBLK); } /* * Cancel a jnewblk which has been been made redundant by frag extension. */ static void cancel_jnewblk(jnewblk, wkhd) struct jnewblk *jnewblk; struct workhead *wkhd; { struct jsegdep *jsegdep; CTR1(KTR_SUJ, "cancel_jnewblk: blkno %jd", jnewblk->jn_blkno); jsegdep = jnewblk->jn_jsegdep; if (jnewblk->jn_jsegdep == NULL || jnewblk->jn_dep == NULL) panic("cancel_jnewblk: Invalid state"); jnewblk->jn_jsegdep = NULL; jnewblk->jn_dep = NULL; jnewblk->jn_state |= GOINGAWAY; if (jnewblk->jn_state & INPROGRESS) { jnewblk->jn_state &= ~INPROGRESS; WORKLIST_REMOVE(&jnewblk->jn_list); jwork_insert(wkhd, jsegdep); } else { free_jsegdep(jsegdep); remove_from_journal(&jnewblk->jn_list); } wake_worklist(&jnewblk->jn_list); WORKLIST_INSERT(wkhd, &jnewblk->jn_list); } static void free_jblkdep(jblkdep) struct jblkdep *jblkdep; { if (jblkdep->jb_list.wk_type == D_JFREEBLK) WORKITEM_FREE(jblkdep, D_JFREEBLK); else if (jblkdep->jb_list.wk_type == D_JTRUNC) WORKITEM_FREE(jblkdep, D_JTRUNC); else panic("free_jblkdep: Unexpected type %s", TYPENAME(jblkdep->jb_list.wk_type)); } /* * Free a single jseg once it is no longer referenced in memory or on * disk. Reclaim journal blocks and dependencies waiting for the segment * to disappear. */ static void free_jseg(jseg, jblocks) struct jseg *jseg; struct jblocks *jblocks; { struct freework *freework; /* * Free freework structures that were lingering to indicate freed * indirect blocks that forced journal write ordering on reallocate. */ while ((freework = LIST_FIRST(&jseg->js_indirs)) != NULL) indirblk_remove(freework); if (jblocks->jb_oldestseg == jseg) jblocks->jb_oldestseg = TAILQ_NEXT(jseg, js_next); TAILQ_REMOVE(&jblocks->jb_segs, jseg, js_next); jblocks_free(jblocks, jseg->js_list.wk_mp, jseg->js_size); KASSERT(LIST_EMPTY(&jseg->js_entries), ("free_jseg: Freed jseg has valid entries.")); WORKITEM_FREE(jseg, D_JSEG); } /* * Free all jsegs that meet the criteria for being reclaimed and update * oldestseg. */ static void free_jsegs(jblocks) struct jblocks *jblocks; { struct jseg *jseg; /* * Free only those jsegs which have none allocated before them to * preserve the journal space ordering. */ while ((jseg = TAILQ_FIRST(&jblocks->jb_segs)) != NULL) { /* * Only reclaim space when nothing depends on this journal * set and another set has written that it is no longer * valid. */ if (jseg->js_refs != 0) { jblocks->jb_oldestseg = jseg; return; } if ((jseg->js_state & ALLCOMPLETE) != ALLCOMPLETE) break; if (jseg->js_seq > jblocks->jb_oldestwrseq) break; /* * We can free jsegs that didn't write entries when * oldestwrseq == js_seq. */ if (jseg->js_seq == jblocks->jb_oldestwrseq && jseg->js_cnt != 0) break; free_jseg(jseg, jblocks); } /* * If we exited the loop above we still must discover the * oldest valid segment. */ if (jseg) for (jseg = jblocks->jb_oldestseg; jseg != NULL; jseg = TAILQ_NEXT(jseg, js_next)) if (jseg->js_refs != 0) break; jblocks->jb_oldestseg = jseg; /* * The journal has no valid records but some jsegs may still be * waiting on oldestwrseq to advance. We force a small record * out to permit these lingering records to be reclaimed. */ if (jblocks->jb_oldestseg == NULL && !TAILQ_EMPTY(&jblocks->jb_segs)) jblocks->jb_needseg = 1; } /* * Release one reference to a jseg and free it if the count reaches 0. This * should eventually reclaim journal space as well. */ static void rele_jseg(jseg) struct jseg *jseg; { KASSERT(jseg->js_refs > 0, ("free_jseg: Invalid refcnt %d", jseg->js_refs)); if (--jseg->js_refs != 0) return; free_jsegs(jseg->js_jblocks); } /* * Release a jsegdep and decrement the jseg count. */ static void free_jsegdep(jsegdep) struct jsegdep *jsegdep; { if (jsegdep->jd_seg) rele_jseg(jsegdep->jd_seg); WORKITEM_FREE(jsegdep, D_JSEGDEP); } /* * Wait for a journal item to make it to disk. Initiate journal processing * if required. */ static int jwait(wk, waitfor) struct worklist *wk; int waitfor; { LOCK_OWNED(VFSTOUFS(wk->wk_mp)); /* * Blocking journal waits cause slow synchronous behavior. Record * stats on the frequency of these blocking operations. */ if (waitfor == MNT_WAIT) { stat_journal_wait++; switch (wk->wk_type) { case D_JREMREF: case D_JMVREF: stat_jwait_filepage++; break; case D_JTRUNC: case D_JFREEBLK: stat_jwait_freeblks++; break; case D_JNEWBLK: stat_jwait_newblk++; break; case D_JADDREF: stat_jwait_inode++; break; default: break; } } /* * If IO has not started we process the journal. We can't mark the * worklist item as IOWAITING because we drop the lock while * processing the journal and the worklist entry may be freed after * this point. The caller may call back in and re-issue the request. */ if ((wk->wk_state & INPROGRESS) == 0) { softdep_process_journal(wk->wk_mp, wk, waitfor); if (waitfor != MNT_WAIT) return (EBUSY); return (0); } if (waitfor != MNT_WAIT) return (EBUSY); wait_worklist(wk, "jwait"); return (0); } /* * Lookup an inodedep based on an inode pointer and set the nlinkdelta as * appropriate. This is a convenience function to reduce duplicate code * for the setup and revert functions below. */ static struct inodedep * inodedep_lookup_ip(ip) struct inode *ip; { struct inodedep *inodedep; KASSERT(ip->i_nlink >= ip->i_effnlink, ("inodedep_lookup_ip: bad delta")); (void) inodedep_lookup(UFSTOVFS(ip->i_ump), ip->i_number, DEPALLOC, &inodedep); inodedep->id_nlinkdelta = ip->i_nlink - ip->i_effnlink; KASSERT((inodedep->id_state & UNLINKED) == 0, ("inode unlinked")); return (inodedep); } /* * Called prior to creating a new inode and linking it to a directory. The * jaddref structure must already be allocated by softdep_setup_inomapdep * and it is discovered here so we can initialize the mode and update * nlinkdelta. */ void softdep_setup_create(dp, ip) struct inode *dp; struct inode *ip; { struct inodedep *inodedep; struct jaddref *jaddref; struct vnode *dvp; KASSERT(MOUNTEDSOFTDEP(UFSTOVFS(dp->i_ump)) != 0, ("softdep_setup_create called on non-softdep filesystem")); KASSERT(ip->i_nlink == 1, ("softdep_setup_create: Invalid link count.")); dvp = ITOV(dp); ACQUIRE_LOCK(dp->i_ump); inodedep = inodedep_lookup_ip(ip); if (DOINGSUJ(dvp)) { jaddref = (struct jaddref *)TAILQ_LAST(&inodedep->id_inoreflst, inoreflst); KASSERT(jaddref != NULL && jaddref->ja_parent == dp->i_number, ("softdep_setup_create: No addref structure present.")); } softdep_prelink(dvp, NULL); FREE_LOCK(dp->i_ump); } /* * Create a jaddref structure to track the addition of a DOTDOT link when * we are reparenting an inode as part of a rename. This jaddref will be * found by softdep_setup_directory_change. Adjusts nlinkdelta for * non-journaling softdep. */ void softdep_setup_dotdot_link(dp, ip) struct inode *dp; struct inode *ip; { struct inodedep *inodedep; struct jaddref *jaddref; struct vnode *dvp; KASSERT(MOUNTEDSOFTDEP(UFSTOVFS(dp->i_ump)) != 0, ("softdep_setup_dotdot_link called on non-softdep filesystem")); dvp = ITOV(dp); jaddref = NULL; /* * We don't set MKDIR_PARENT as this is not tied to a mkdir and * is used as a normal link would be. */ if (DOINGSUJ(dvp)) jaddref = newjaddref(ip, dp->i_number, DOTDOT_OFFSET, dp->i_effnlink - 1, dp->i_mode); ACQUIRE_LOCK(dp->i_ump); inodedep = inodedep_lookup_ip(dp); if (jaddref) TAILQ_INSERT_TAIL(&inodedep->id_inoreflst, &jaddref->ja_ref, if_deps); softdep_prelink(dvp, ITOV(ip)); FREE_LOCK(dp->i_ump); } /* * Create a jaddref structure to track a new link to an inode. The directory * offset is not known until softdep_setup_directory_add or * softdep_setup_directory_change. Adjusts nlinkdelta for non-journaling * softdep. */ void softdep_setup_link(dp, ip) struct inode *dp; struct inode *ip; { struct inodedep *inodedep; struct jaddref *jaddref; struct vnode *dvp; KASSERT(MOUNTEDSOFTDEP(UFSTOVFS(dp->i_ump)) != 0, ("softdep_setup_link called on non-softdep filesystem")); dvp = ITOV(dp); jaddref = NULL; if (DOINGSUJ(dvp)) jaddref = newjaddref(dp, ip->i_number, 0, ip->i_effnlink - 1, ip->i_mode); ACQUIRE_LOCK(dp->i_ump); inodedep = inodedep_lookup_ip(ip); if (jaddref) TAILQ_INSERT_TAIL(&inodedep->id_inoreflst, &jaddref->ja_ref, if_deps); softdep_prelink(dvp, ITOV(ip)); FREE_LOCK(dp->i_ump); } /* * Called to create the jaddref structures to track . and .. references as * well as lookup and further initialize the incomplete jaddref created * by softdep_setup_inomapdep when the inode was allocated. Adjusts * nlinkdelta for non-journaling softdep. */ void softdep_setup_mkdir(dp, ip) struct inode *dp; struct inode *ip; { struct inodedep *inodedep; struct jaddref *dotdotaddref; struct jaddref *dotaddref; struct jaddref *jaddref; struct vnode *dvp; KASSERT(MOUNTEDSOFTDEP(UFSTOVFS(dp->i_ump)) != 0, ("softdep_setup_mkdir called on non-softdep filesystem")); dvp = ITOV(dp); dotaddref = dotdotaddref = NULL; if (DOINGSUJ(dvp)) { dotaddref = newjaddref(ip, ip->i_number, DOT_OFFSET, 1, ip->i_mode); dotaddref->ja_state |= MKDIR_BODY; dotdotaddref = newjaddref(ip, dp->i_number, DOTDOT_OFFSET, dp->i_effnlink - 1, dp->i_mode); dotdotaddref->ja_state |= MKDIR_PARENT; } ACQUIRE_LOCK(dp->i_ump); inodedep = inodedep_lookup_ip(ip); if (DOINGSUJ(dvp)) { jaddref = (struct jaddref *)TAILQ_LAST(&inodedep->id_inoreflst, inoreflst); KASSERT(jaddref != NULL, ("softdep_setup_mkdir: No addref structure present.")); KASSERT(jaddref->ja_parent == dp->i_number, ("softdep_setup_mkdir: bad parent %ju", (uintmax_t)jaddref->ja_parent)); TAILQ_INSERT_BEFORE(&jaddref->ja_ref, &dotaddref->ja_ref, if_deps); } inodedep = inodedep_lookup_ip(dp); if (DOINGSUJ(dvp)) TAILQ_INSERT_TAIL(&inodedep->id_inoreflst, &dotdotaddref->ja_ref, if_deps); softdep_prelink(ITOV(dp), NULL); FREE_LOCK(dp->i_ump); } /* * Called to track nlinkdelta of the inode and parent directories prior to * unlinking a directory. */ void softdep_setup_rmdir(dp, ip) struct inode *dp; struct inode *ip; { struct vnode *dvp; KASSERT(MOUNTEDSOFTDEP(UFSTOVFS(dp->i_ump)) != 0, ("softdep_setup_rmdir called on non-softdep filesystem")); dvp = ITOV(dp); ACQUIRE_LOCK(dp->i_ump); (void) inodedep_lookup_ip(ip); (void) inodedep_lookup_ip(dp); softdep_prelink(dvp, ITOV(ip)); FREE_LOCK(dp->i_ump); } /* * Called to track nlinkdelta of the inode and parent directories prior to * unlink. */ void softdep_setup_unlink(dp, ip) struct inode *dp; struct inode *ip; { struct vnode *dvp; KASSERT(MOUNTEDSOFTDEP(UFSTOVFS(dp->i_ump)) != 0, ("softdep_setup_unlink called on non-softdep filesystem")); dvp = ITOV(dp); ACQUIRE_LOCK(dp->i_ump); (void) inodedep_lookup_ip(ip); (void) inodedep_lookup_ip(dp); softdep_prelink(dvp, ITOV(ip)); FREE_LOCK(dp->i_ump); } /* * Called to release the journal structures created by a failed non-directory * creation. Adjusts nlinkdelta for non-journaling softdep. */ void softdep_revert_create(dp, ip) struct inode *dp; struct inode *ip; { struct inodedep *inodedep; struct jaddref *jaddref; struct vnode *dvp; KASSERT(MOUNTEDSOFTDEP(UFSTOVFS(dp->i_ump)) != 0, ("softdep_revert_create called on non-softdep filesystem")); dvp = ITOV(dp); ACQUIRE_LOCK(dp->i_ump); inodedep = inodedep_lookup_ip(ip); if (DOINGSUJ(dvp)) { jaddref = (struct jaddref *)TAILQ_LAST(&inodedep->id_inoreflst, inoreflst); KASSERT(jaddref->ja_parent == dp->i_number, ("softdep_revert_create: addref parent mismatch")); cancel_jaddref(jaddref, inodedep, &inodedep->id_inowait); } FREE_LOCK(dp->i_ump); } /* * Called to release the journal structures created by a failed link * addition. Adjusts nlinkdelta for non-journaling softdep. */ void softdep_revert_link(dp, ip) struct inode *dp; struct inode *ip; { struct inodedep *inodedep; struct jaddref *jaddref; struct vnode *dvp; KASSERT(MOUNTEDSOFTDEP(UFSTOVFS(dp->i_ump)) != 0, ("softdep_revert_link called on non-softdep filesystem")); dvp = ITOV(dp); ACQUIRE_LOCK(dp->i_ump); inodedep = inodedep_lookup_ip(ip); if (DOINGSUJ(dvp)) { jaddref = (struct jaddref *)TAILQ_LAST(&inodedep->id_inoreflst, inoreflst); KASSERT(jaddref->ja_parent == dp->i_number, ("softdep_revert_link: addref parent mismatch")); cancel_jaddref(jaddref, inodedep, &inodedep->id_inowait); } FREE_LOCK(dp->i_ump); } /* * Called to release the journal structures created by a failed mkdir * attempt. Adjusts nlinkdelta for non-journaling softdep. */ void softdep_revert_mkdir(dp, ip) struct inode *dp; struct inode *ip; { struct inodedep *inodedep; struct jaddref *jaddref; struct jaddref *dotaddref; struct vnode *dvp; KASSERT(MOUNTEDSOFTDEP(UFSTOVFS(dp->i_ump)) != 0, ("softdep_revert_mkdir called on non-softdep filesystem")); dvp = ITOV(dp); ACQUIRE_LOCK(dp->i_ump); inodedep = inodedep_lookup_ip(dp); if (DOINGSUJ(dvp)) { jaddref = (struct jaddref *)TAILQ_LAST(&inodedep->id_inoreflst, inoreflst); KASSERT(jaddref->ja_parent == ip->i_number, ("softdep_revert_mkdir: dotdot addref parent mismatch")); cancel_jaddref(jaddref, inodedep, &inodedep->id_inowait); } inodedep = inodedep_lookup_ip(ip); if (DOINGSUJ(dvp)) { jaddref = (struct jaddref *)TAILQ_LAST(&inodedep->id_inoreflst, inoreflst); KASSERT(jaddref->ja_parent == dp->i_number, ("softdep_revert_mkdir: addref parent mismatch")); dotaddref = (struct jaddref *)TAILQ_PREV(&jaddref->ja_ref, inoreflst, if_deps); cancel_jaddref(jaddref, inodedep, &inodedep->id_inowait); KASSERT(dotaddref->ja_parent == ip->i_number, ("softdep_revert_mkdir: dot addref parent mismatch")); cancel_jaddref(dotaddref, inodedep, &inodedep->id_inowait); } FREE_LOCK(dp->i_ump); } /* * Called to correct nlinkdelta after a failed rmdir. */ void softdep_revert_rmdir(dp, ip) struct inode *dp; struct inode *ip; { KASSERT(MOUNTEDSOFTDEP(UFSTOVFS(dp->i_ump)) != 0, ("softdep_revert_rmdir called on non-softdep filesystem")); ACQUIRE_LOCK(dp->i_ump); (void) inodedep_lookup_ip(ip); (void) inodedep_lookup_ip(dp); FREE_LOCK(dp->i_ump); } /* * Protecting the freemaps (or bitmaps). * * To eliminate the need to execute fsck before mounting a filesystem * after a power failure, one must (conservatively) guarantee that the * on-disk copy of the bitmaps never indicate that a live inode or block is * free. So, when a block or inode is allocated, the bitmap should be * updated (on disk) before any new pointers. When a block or inode is * freed, the bitmap should not be updated until all pointers have been * reset. The latter dependency is handled by the delayed de-allocation * approach described below for block and inode de-allocation. The former * dependency is handled by calling the following procedure when a block or * inode is allocated. When an inode is allocated an "inodedep" is created * with its DEPCOMPLETE flag cleared until its bitmap is written to disk. * Each "inodedep" is also inserted into the hash indexing structure so * that any additional link additions can be made dependent on the inode * allocation. * * The ufs filesystem maintains a number of free block counts (e.g., per * cylinder group, per cylinder and per pair) * in addition to the bitmaps. These counts are used to improve efficiency * during allocation and therefore must be consistent with the bitmaps. * There is no convenient way to guarantee post-crash consistency of these * counts with simple update ordering, for two main reasons: (1) The counts * and bitmaps for a single cylinder group block are not in the same disk * sector. If a disk write is interrupted (e.g., by power failure), one may * be written and the other not. (2) Some of the counts are located in the * superblock rather than the cylinder group block. So, we focus our soft * updates implementation on protecting the bitmaps. When mounting a * filesystem, we recompute the auxiliary counts from the bitmaps. */ /* * Called just after updating the cylinder group block to allocate an inode. */ void softdep_setup_inomapdep(bp, ip, newinum, mode) struct buf *bp; /* buffer for cylgroup block with inode map */ struct inode *ip; /* inode related to allocation */ ino_t newinum; /* new inode number being allocated */ int mode; { struct inodedep *inodedep; struct bmsafemap *bmsafemap; struct jaddref *jaddref; struct mount *mp; struct fs *fs; mp = UFSTOVFS(ip->i_ump); KASSERT(MOUNTEDSOFTDEP(mp) != 0, ("softdep_setup_inomapdep called on non-softdep filesystem")); fs = ip->i_ump->um_fs; jaddref = NULL; /* * Allocate the journal reference add structure so that the bitmap * can be dependent on it. */ if (MOUNTEDSUJ(mp)) { jaddref = newjaddref(ip, newinum, 0, 0, mode); jaddref->ja_state |= NEWBLOCK; } /* * Create a dependency for the newly allocated inode. * Panic if it already exists as something is seriously wrong. * Otherwise add it to the dependency list for the buffer holding * the cylinder group map from which it was allocated. * * We have to preallocate a bmsafemap entry in case it is needed * in bmsafemap_lookup since once we allocate the inodedep, we * have to finish initializing it before we can FREE_LOCK(). * By preallocating, we avoid FREE_LOCK() while doing a malloc * in bmsafemap_lookup. We cannot call bmsafemap_lookup before * creating the inodedep as it can be freed during the time * that we FREE_LOCK() while allocating the inodedep. We must * call workitem_alloc() before entering the locked section as * it also acquires the lock and we must avoid trying doing so * recursively. */ bmsafemap = malloc(sizeof(struct bmsafemap), M_BMSAFEMAP, M_SOFTDEP_FLAGS); workitem_alloc(&bmsafemap->sm_list, D_BMSAFEMAP, mp); ACQUIRE_LOCK(ip->i_ump); if ((inodedep_lookup(mp, newinum, DEPALLOC, &inodedep))) panic("softdep_setup_inomapdep: dependency %p for new" "inode already exists", inodedep); bmsafemap = bmsafemap_lookup(mp, bp, ino_to_cg(fs, newinum), bmsafemap); if (jaddref) { LIST_INSERT_HEAD(&bmsafemap->sm_jaddrefhd, jaddref, ja_bmdeps); TAILQ_INSERT_TAIL(&inodedep->id_inoreflst, &jaddref->ja_ref, if_deps); } else { inodedep->id_state |= ONDEPLIST; LIST_INSERT_HEAD(&bmsafemap->sm_inodedephd, inodedep, id_deps); } inodedep->id_bmsafemap = bmsafemap; inodedep->id_state &= ~DEPCOMPLETE; FREE_LOCK(ip->i_ump); } /* * Called just after updating the cylinder group block to * allocate block or fragment. */ void softdep_setup_blkmapdep(bp, mp, newblkno, frags, oldfrags) struct buf *bp; /* buffer for cylgroup block with block map */ struct mount *mp; /* filesystem doing allocation */ ufs2_daddr_t newblkno; /* number of newly allocated block */ int frags; /* Number of fragments. */ int oldfrags; /* Previous number of fragments for extend. */ { struct newblk *newblk; struct bmsafemap *bmsafemap; struct jnewblk *jnewblk; struct ufsmount *ump; struct fs *fs; KASSERT(MOUNTEDSOFTDEP(mp) != 0, ("softdep_setup_blkmapdep called on non-softdep filesystem")); ump = VFSTOUFS(mp); fs = ump->um_fs; jnewblk = NULL; /* * Create a dependency for the newly allocated block. * Add it to the dependency list for the buffer holding * the cylinder group map from which it was allocated. */ if (MOUNTEDSUJ(mp)) { jnewblk = malloc(sizeof(*jnewblk), M_JNEWBLK, M_SOFTDEP_FLAGS); workitem_alloc(&jnewblk->jn_list, D_JNEWBLK, mp); jnewblk->jn_jsegdep = newjsegdep(&jnewblk->jn_list); jnewblk->jn_state = ATTACHED; jnewblk->jn_blkno = newblkno; jnewblk->jn_frags = frags; jnewblk->jn_oldfrags = oldfrags; #ifdef SUJ_DEBUG { struct cg *cgp; uint8_t *blksfree; long bno; int i; cgp = (struct cg *)bp->b_data; blksfree = cg_blksfree(cgp); bno = dtogd(fs, jnewblk->jn_blkno); for (i = jnewblk->jn_oldfrags; i < jnewblk->jn_frags; i++) { if (isset(blksfree, bno + i)) panic("softdep_setup_blkmapdep: " "free fragment %d from %d-%d " "state 0x%X dep %p", i, jnewblk->jn_oldfrags, jnewblk->jn_frags, jnewblk->jn_state, jnewblk->jn_dep); } } #endif } CTR3(KTR_SUJ, "softdep_setup_blkmapdep: blkno %jd frags %d oldfrags %d", newblkno, frags, oldfrags); ACQUIRE_LOCK(ump); if (newblk_lookup(mp, newblkno, DEPALLOC, &newblk) != 0) panic("softdep_setup_blkmapdep: found block"); newblk->nb_bmsafemap = bmsafemap = bmsafemap_lookup(mp, bp, dtog(fs, newblkno), NULL); if (jnewblk) { jnewblk->jn_dep = (struct worklist *)newblk; LIST_INSERT_HEAD(&bmsafemap->sm_jnewblkhd, jnewblk, jn_deps); } else { newblk->nb_state |= ONDEPLIST; LIST_INSERT_HEAD(&bmsafemap->sm_newblkhd, newblk, nb_deps); } newblk->nb_bmsafemap = bmsafemap; newblk->nb_jnewblk = jnewblk; FREE_LOCK(ump); } #define BMSAFEMAP_HASH(ump, cg) \ (&(ump)->bmsafemap_hashtbl[(cg) & (ump)->bmsafemap_hash_size]) static int bmsafemap_find(bmsafemaphd, cg, bmsafemapp) struct bmsafemap_hashhead *bmsafemaphd; int cg; struct bmsafemap **bmsafemapp; { struct bmsafemap *bmsafemap; LIST_FOREACH(bmsafemap, bmsafemaphd, sm_hash) if (bmsafemap->sm_cg == cg) break; if (bmsafemap) { *bmsafemapp = bmsafemap; return (1); } *bmsafemapp = NULL; return (0); } /* * Find the bmsafemap associated with a cylinder group buffer. * If none exists, create one. The buffer must be locked when * this routine is called and this routine must be called with * the softdep lock held. To avoid giving up the lock while * allocating a new bmsafemap, a preallocated bmsafemap may be * provided. If it is provided but not needed, it is freed. */ static struct bmsafemap * bmsafemap_lookup(mp, bp, cg, newbmsafemap) struct mount *mp; struct buf *bp; int cg; struct bmsafemap *newbmsafemap; { struct bmsafemap_hashhead *bmsafemaphd; struct bmsafemap *bmsafemap, *collision; struct worklist *wk; struct ufsmount *ump; ump = VFSTOUFS(mp); LOCK_OWNED(ump); KASSERT(bp != NULL, ("bmsafemap_lookup: missing buffer")); LIST_FOREACH(wk, &bp->b_dep, wk_list) { if (wk->wk_type == D_BMSAFEMAP) { if (newbmsafemap) WORKITEM_FREE(newbmsafemap, D_BMSAFEMAP); return (WK_BMSAFEMAP(wk)); } } bmsafemaphd = BMSAFEMAP_HASH(ump, cg); if (bmsafemap_find(bmsafemaphd, cg, &bmsafemap) == 1) { if (newbmsafemap) WORKITEM_FREE(newbmsafemap, D_BMSAFEMAP); return (bmsafemap); } if (newbmsafemap) { bmsafemap = newbmsafemap; } else { FREE_LOCK(ump); bmsafemap = malloc(sizeof(struct bmsafemap), M_BMSAFEMAP, M_SOFTDEP_FLAGS); workitem_alloc(&bmsafemap->sm_list, D_BMSAFEMAP, mp); ACQUIRE_LOCK(ump); } bmsafemap->sm_buf = bp; LIST_INIT(&bmsafemap->sm_inodedephd); LIST_INIT(&bmsafemap->sm_inodedepwr); LIST_INIT(&bmsafemap->sm_newblkhd); LIST_INIT(&bmsafemap->sm_newblkwr); LIST_INIT(&bmsafemap->sm_jaddrefhd); LIST_INIT(&bmsafemap->sm_jnewblkhd); LIST_INIT(&bmsafemap->sm_freehd); LIST_INIT(&bmsafemap->sm_freewr); if (bmsafemap_find(bmsafemaphd, cg, &collision) == 1) { WORKITEM_FREE(bmsafemap, D_BMSAFEMAP); return (collision); } bmsafemap->sm_cg = cg; LIST_INSERT_HEAD(bmsafemaphd, bmsafemap, sm_hash); LIST_INSERT_HEAD(&ump->softdep_dirtycg, bmsafemap, sm_next); WORKLIST_INSERT(&bp->b_dep, &bmsafemap->sm_list); return (bmsafemap); } /* * Direct block allocation dependencies. * * When a new block is allocated, the corresponding disk locations must be * initialized (with zeros or new data) before the on-disk inode points to * them. Also, the freemap from which the block was allocated must be * updated (on disk) before the inode's pointer. These two dependencies are * independent of each other and are needed for all file blocks and indirect * blocks that are pointed to directly by the inode. Just before the * "in-core" version of the inode is updated with a newly allocated block * number, a procedure (below) is called to setup allocation dependency * structures. These structures are removed when the corresponding * dependencies are satisfied or when the block allocation becomes obsolete * (i.e., the file is deleted, the block is de-allocated, or the block is a * fragment that gets upgraded). All of these cases are handled in * procedures described later. * * When a file extension causes a fragment to be upgraded, either to a larger * fragment or to a full block, the on-disk location may change (if the * previous fragment could not simply be extended). In this case, the old * fragment must be de-allocated, but not until after the inode's pointer has * been updated. In most cases, this is handled by later procedures, which * will construct a "freefrag" structure to be added to the workitem queue * when the inode update is complete (or obsolete). The main exception to * this is when an allocation occurs while a pending allocation dependency * (for the same block pointer) remains. This case is handled in the main * allocation dependency setup procedure by immediately freeing the * unreferenced fragments. */ void softdep_setup_allocdirect(ip, off, newblkno, oldblkno, newsize, oldsize, bp) struct inode *ip; /* inode to which block is being added */ ufs_lbn_t off; /* block pointer within inode */ ufs2_daddr_t newblkno; /* disk block number being added */ ufs2_daddr_t oldblkno; /* previous block number, 0 unless frag */ long newsize; /* size of new block */ long oldsize; /* size of new block */ struct buf *bp; /* bp for allocated block */ { struct allocdirect *adp, *oldadp; struct allocdirectlst *adphead; struct freefrag *freefrag; struct inodedep *inodedep; struct pagedep *pagedep; struct jnewblk *jnewblk; struct newblk *newblk; struct mount *mp; ufs_lbn_t lbn; lbn = bp->b_lblkno; mp = UFSTOVFS(ip->i_ump); KASSERT(MOUNTEDSOFTDEP(mp) != 0, ("softdep_setup_allocdirect called on non-softdep filesystem")); if (oldblkno && oldblkno != newblkno) freefrag = newfreefrag(ip, oldblkno, oldsize, lbn); else freefrag = NULL; CTR6(KTR_SUJ, "softdep_setup_allocdirect: ino %d blkno %jd oldblkno %jd " "off %jd newsize %ld oldsize %d", ip->i_number, newblkno, oldblkno, off, newsize, oldsize); ACQUIRE_LOCK(ip->i_ump); if (off >= NDADDR) { if (lbn > 0) panic("softdep_setup_allocdirect: bad lbn %jd, off %jd", lbn, off); /* allocating an indirect block */ if (oldblkno != 0) panic("softdep_setup_allocdirect: non-zero indir"); } else { if (off != lbn) panic("softdep_setup_allocdirect: lbn %jd != off %jd", lbn, off); /* * Allocating a direct block. * * If we are allocating a directory block, then we must * allocate an associated pagedep to track additions and * deletions. */ if ((ip->i_mode & IFMT) == IFDIR) pagedep_lookup(mp, bp, ip->i_number, off, DEPALLOC, &pagedep); } if (newblk_lookup(mp, newblkno, 0, &newblk) == 0) panic("softdep_setup_allocdirect: lost block"); KASSERT(newblk->nb_list.wk_type == D_NEWBLK, ("softdep_setup_allocdirect: newblk already initialized")); /* * Convert the newblk to an allocdirect. */ WORKITEM_REASSIGN(newblk, D_ALLOCDIRECT); adp = (struct allocdirect *)newblk; newblk->nb_freefrag = freefrag; adp->ad_offset = off; adp->ad_oldblkno = oldblkno; adp->ad_newsize = newsize; adp->ad_oldsize = oldsize; /* * Finish initializing the journal. */ if ((jnewblk = newblk->nb_jnewblk) != NULL) { jnewblk->jn_ino = ip->i_number; jnewblk->jn_lbn = lbn; add_to_journal(&jnewblk->jn_list); } if (freefrag && freefrag->ff_jdep != NULL && freefrag->ff_jdep->wk_type == D_JFREEFRAG) add_to_journal(freefrag->ff_jdep); inodedep_lookup(mp, ip->i_number, DEPALLOC, &inodedep); adp->ad_inodedep = inodedep; WORKLIST_INSERT(&bp->b_dep, &newblk->nb_list); /* * The list of allocdirects must be kept in sorted and ascending * order so that the rollback routines can quickly determine the * first uncommitted block (the size of the file stored on disk * ends at the end of the lowest committed fragment, or if there * are no fragments, at the end of the highest committed block). * Since files generally grow, the typical case is that the new * block is to be added at the end of the list. We speed this * special case by checking against the last allocdirect in the * list before laboriously traversing the list looking for the * insertion point. */ adphead = &inodedep->id_newinoupdt; oldadp = TAILQ_LAST(adphead, allocdirectlst); if (oldadp == NULL || oldadp->ad_offset <= off) { /* insert at end of list */ TAILQ_INSERT_TAIL(adphead, adp, ad_next); if (oldadp != NULL && oldadp->ad_offset == off) allocdirect_merge(adphead, adp, oldadp); FREE_LOCK(ip->i_ump); return; } TAILQ_FOREACH(oldadp, adphead, ad_next) { if (oldadp->ad_offset >= off) break; } if (oldadp == NULL) panic("softdep_setup_allocdirect: lost entry"); /* insert in middle of list */ TAILQ_INSERT_BEFORE(oldadp, adp, ad_next); if (oldadp->ad_offset == off) allocdirect_merge(adphead, adp, oldadp); FREE_LOCK(ip->i_ump); } /* * Merge a newer and older journal record to be stored either in a * newblock or freefrag. This handles aggregating journal records for * fragment allocation into a second record as well as replacing a * journal free with an aborted journal allocation. A segment for the * oldest record will be placed on wkhd if it has been written. If not * the segment for the newer record will suffice. */ static struct worklist * jnewblk_merge(new, old, wkhd) struct worklist *new; struct worklist *old; struct workhead *wkhd; { struct jnewblk *njnewblk; struct jnewblk *jnewblk; /* Handle NULLs to simplify callers. */ if (new == NULL) return (old); if (old == NULL) return (new); /* Replace a jfreefrag with a jnewblk. */ if (new->wk_type == D_JFREEFRAG) { if (WK_JNEWBLK(old)->jn_blkno != WK_JFREEFRAG(new)->fr_blkno) panic("jnewblk_merge: blkno mismatch: %p, %p", old, new); cancel_jfreefrag(WK_JFREEFRAG(new)); return (old); } if (old->wk_type != D_JNEWBLK || new->wk_type != D_JNEWBLK) panic("jnewblk_merge: Bad type: old %d new %d\n", old->wk_type, new->wk_type); /* * Handle merging of two jnewblk records that describe * different sets of fragments in the same block. */ jnewblk = WK_JNEWBLK(old); njnewblk = WK_JNEWBLK(new); if (jnewblk->jn_blkno != njnewblk->jn_blkno) panic("jnewblk_merge: Merging disparate blocks."); /* * The record may be rolled back in the cg. */ if (jnewblk->jn_state & UNDONE) { jnewblk->jn_state &= ~UNDONE; njnewblk->jn_state |= UNDONE; njnewblk->jn_state &= ~ATTACHED; } /* * We modify the newer addref and free the older so that if neither * has been written the most up-to-date copy will be on disk. If * both have been written but rolled back we only temporarily need * one of them to fix the bits when the cg write completes. */ jnewblk->jn_state |= ATTACHED | COMPLETE; njnewblk->jn_oldfrags = jnewblk->jn_oldfrags; cancel_jnewblk(jnewblk, wkhd); WORKLIST_REMOVE(&jnewblk->jn_list); free_jnewblk(jnewblk); return (new); } /* * Replace an old allocdirect dependency with a newer one. * This routine must be called with splbio interrupts blocked. */ static void allocdirect_merge(adphead, newadp, oldadp) struct allocdirectlst *adphead; /* head of list holding allocdirects */ struct allocdirect *newadp; /* allocdirect being added */ struct allocdirect *oldadp; /* existing allocdirect being checked */ { struct worklist *wk; struct freefrag *freefrag; freefrag = NULL; LOCK_OWNED(VFSTOUFS(newadp->ad_list.wk_mp)); if (newadp->ad_oldblkno != oldadp->ad_newblkno || newadp->ad_oldsize != oldadp->ad_newsize || newadp->ad_offset >= NDADDR) panic("%s %jd != new %jd || old size %ld != new %ld", "allocdirect_merge: old blkno", (intmax_t)newadp->ad_oldblkno, (intmax_t)oldadp->ad_newblkno, newadp->ad_oldsize, oldadp->ad_newsize); newadp->ad_oldblkno = oldadp->ad_oldblkno; newadp->ad_oldsize = oldadp->ad_oldsize; /* * If the old dependency had a fragment to free or had never * previously had a block allocated, then the new dependency * can immediately post its freefrag and adopt the old freefrag. * This action is done by swapping the freefrag dependencies. * The new dependency gains the old one's freefrag, and the * old one gets the new one and then immediately puts it on * the worklist when it is freed by free_newblk. It is * not possible to do this swap when the old dependency had a * non-zero size but no previous fragment to free. This condition * arises when the new block is an extension of the old block. * Here, the first part of the fragment allocated to the new * dependency is part of the block currently claimed on disk by * the old dependency, so cannot legitimately be freed until the * conditions for the new dependency are fulfilled. */ freefrag = newadp->ad_freefrag; if (oldadp->ad_freefrag != NULL || oldadp->ad_oldblkno == 0) { newadp->ad_freefrag = oldadp->ad_freefrag; oldadp->ad_freefrag = freefrag; } /* * If we are tracking a new directory-block allocation, * move it from the old allocdirect to the new allocdirect. */ if ((wk = LIST_FIRST(&oldadp->ad_newdirblk)) != NULL) { WORKLIST_REMOVE(wk); if (!LIST_EMPTY(&oldadp->ad_newdirblk)) panic("allocdirect_merge: extra newdirblk"); WORKLIST_INSERT(&newadp->ad_newdirblk, wk); } TAILQ_REMOVE(adphead, oldadp, ad_next); /* * We need to move any journal dependencies over to the freefrag * that releases this block if it exists. Otherwise we are * extending an existing block and we'll wait until that is * complete to release the journal space and extend the * new journal to cover this old space as well. */ if (freefrag == NULL) { if (oldadp->ad_newblkno != newadp->ad_newblkno) panic("allocdirect_merge: %jd != %jd", oldadp->ad_newblkno, newadp->ad_newblkno); newadp->ad_block.nb_jnewblk = (struct jnewblk *) jnewblk_merge(&newadp->ad_block.nb_jnewblk->jn_list, &oldadp->ad_block.nb_jnewblk->jn_list, &newadp->ad_block.nb_jwork); oldadp->ad_block.nb_jnewblk = NULL; cancel_newblk(&oldadp->ad_block, NULL, &newadp->ad_block.nb_jwork); } else { wk = (struct worklist *) cancel_newblk(&oldadp->ad_block, &freefrag->ff_list, &freefrag->ff_jwork); freefrag->ff_jdep = jnewblk_merge(freefrag->ff_jdep, wk, &freefrag->ff_jwork); } free_newblk(&oldadp->ad_block); } /* * Allocate a jfreefrag structure to journal a single block free. */ static struct jfreefrag * newjfreefrag(freefrag, ip, blkno, size, lbn) struct freefrag *freefrag; struct inode *ip; ufs2_daddr_t blkno; long size; ufs_lbn_t lbn; { struct jfreefrag *jfreefrag; struct fs *fs; fs = ip->i_fs; jfreefrag = malloc(sizeof(struct jfreefrag), M_JFREEFRAG, M_SOFTDEP_FLAGS); workitem_alloc(&jfreefrag->fr_list, D_JFREEFRAG, UFSTOVFS(ip->i_ump)); jfreefrag->fr_jsegdep = newjsegdep(&jfreefrag->fr_list); jfreefrag->fr_state = ATTACHED | DEPCOMPLETE; jfreefrag->fr_ino = ip->i_number; jfreefrag->fr_lbn = lbn; jfreefrag->fr_blkno = blkno; jfreefrag->fr_frags = numfrags(fs, size); jfreefrag->fr_freefrag = freefrag; return (jfreefrag); } /* * Allocate a new freefrag structure. */ static struct freefrag * newfreefrag(ip, blkno, size, lbn) struct inode *ip; ufs2_daddr_t blkno; long size; ufs_lbn_t lbn; { struct freefrag *freefrag; struct fs *fs; CTR4(KTR_SUJ, "newfreefrag: ino %d blkno %jd size %ld lbn %jd", ip->i_number, blkno, size, lbn); fs = ip->i_fs; if (fragnum(fs, blkno) + numfrags(fs, size) > fs->fs_frag) panic("newfreefrag: frag size"); freefrag = malloc(sizeof(struct freefrag), M_FREEFRAG, M_SOFTDEP_FLAGS); workitem_alloc(&freefrag->ff_list, D_FREEFRAG, UFSTOVFS(ip->i_ump)); freefrag->ff_state = ATTACHED; LIST_INIT(&freefrag->ff_jwork); freefrag->ff_inum = ip->i_number; freefrag->ff_vtype = ITOV(ip)->v_type; freefrag->ff_blkno = blkno; freefrag->ff_fragsize = size; if (MOUNTEDSUJ(UFSTOVFS(ip->i_ump))) { freefrag->ff_jdep = (struct worklist *) newjfreefrag(freefrag, ip, blkno, size, lbn); } else { freefrag->ff_state |= DEPCOMPLETE; freefrag->ff_jdep = NULL; } return (freefrag); } /* * This workitem de-allocates fragments that were replaced during * file block allocation. */ static void handle_workitem_freefrag(freefrag) struct freefrag *freefrag; { struct ufsmount *ump = VFSTOUFS(freefrag->ff_list.wk_mp); struct workhead wkhd; CTR3(KTR_SUJ, "handle_workitem_freefrag: ino %d blkno %jd size %ld", freefrag->ff_inum, freefrag->ff_blkno, freefrag->ff_fragsize); /* * It would be illegal to add new completion items to the * freefrag after it was schedule to be done so it must be * safe to modify the list head here. */ LIST_INIT(&wkhd); ACQUIRE_LOCK(ump); LIST_SWAP(&freefrag->ff_jwork, &wkhd, worklist, wk_list); /* * If the journal has not been written we must cancel it here. */ if (freefrag->ff_jdep) { if (freefrag->ff_jdep->wk_type != D_JNEWBLK) panic("handle_workitem_freefrag: Unexpected type %d\n", freefrag->ff_jdep->wk_type); cancel_jnewblk(WK_JNEWBLK(freefrag->ff_jdep), &wkhd); } FREE_LOCK(ump); ffs_blkfree(ump, ump->um_fs, ump->um_devvp, freefrag->ff_blkno, freefrag->ff_fragsize, freefrag->ff_inum, freefrag->ff_vtype, &wkhd); ACQUIRE_LOCK(ump); WORKITEM_FREE(freefrag, D_FREEFRAG); FREE_LOCK(ump); } /* * Set up a dependency structure for an external attributes data block. * This routine follows much of the structure of softdep_setup_allocdirect. * See the description of softdep_setup_allocdirect above for details. */ void softdep_setup_allocext(ip, off, newblkno, oldblkno, newsize, oldsize, bp) struct inode *ip; ufs_lbn_t off; ufs2_daddr_t newblkno; ufs2_daddr_t oldblkno; long newsize; long oldsize; struct buf *bp; { struct allocdirect *adp, *oldadp; struct allocdirectlst *adphead; struct freefrag *freefrag; struct inodedep *inodedep; struct jnewblk *jnewblk; struct newblk *newblk; struct mount *mp; ufs_lbn_t lbn; mp = UFSTOVFS(ip->i_ump); KASSERT(MOUNTEDSOFTDEP(mp) != 0, ("softdep_setup_allocext called on non-softdep filesystem")); KASSERT(off < NXADDR, ("softdep_setup_allocext: lbn %lld > NXADDR", (long long)off)); lbn = bp->b_lblkno; if (oldblkno && oldblkno != newblkno) freefrag = newfreefrag(ip, oldblkno, oldsize, lbn); else freefrag = NULL; ACQUIRE_LOCK(ip->i_ump); if (newblk_lookup(mp, newblkno, 0, &newblk) == 0) panic("softdep_setup_allocext: lost block"); KASSERT(newblk->nb_list.wk_type == D_NEWBLK, ("softdep_setup_allocext: newblk already initialized")); /* * Convert the newblk to an allocdirect. */ WORKITEM_REASSIGN(newblk, D_ALLOCDIRECT); adp = (struct allocdirect *)newblk; newblk->nb_freefrag = freefrag; adp->ad_offset = off; adp->ad_oldblkno = oldblkno; adp->ad_newsize = newsize; adp->ad_oldsize = oldsize; adp->ad_state |= EXTDATA; /* * Finish initializing the journal. */ if ((jnewblk = newblk->nb_jnewblk) != NULL) { jnewblk->jn_ino = ip->i_number; jnewblk->jn_lbn = lbn; add_to_journal(&jnewblk->jn_list); } if (freefrag && freefrag->ff_jdep != NULL && freefrag->ff_jdep->wk_type == D_JFREEFRAG) add_to_journal(freefrag->ff_jdep); inodedep_lookup(mp, ip->i_number, DEPALLOC, &inodedep); adp->ad_inodedep = inodedep; WORKLIST_INSERT(&bp->b_dep, &newblk->nb_list); /* * The list of allocdirects must be kept in sorted and ascending * order so that the rollback routines can quickly determine the * first uncommitted block (the size of the file stored on disk * ends at the end of the lowest committed fragment, or if there * are no fragments, at the end of the highest committed block). * Since files generally grow, the typical case is that the new * block is to be added at the end of the list. We speed this * special case by checking against the last allocdirect in the * list before laboriously traversing the list looking for the * insertion point. */ adphead = &inodedep->id_newextupdt; oldadp = TAILQ_LAST(adphead, allocdirectlst); if (oldadp == NULL || oldadp->ad_offset <= off) { /* insert at end of list */ TAILQ_INSERT_TAIL(adphead, adp, ad_next); if (oldadp != NULL && oldadp->ad_offset == off) allocdirect_merge(adphead, adp, oldadp); FREE_LOCK(ip->i_ump); return; } TAILQ_FOREACH(oldadp, adphead, ad_next) { if (oldadp->ad_offset >= off) break; } if (oldadp == NULL) panic("softdep_setup_allocext: lost entry"); /* insert in middle of list */ TAILQ_INSERT_BEFORE(oldadp, adp, ad_next); if (oldadp->ad_offset == off) allocdirect_merge(adphead, adp, oldadp); FREE_LOCK(ip->i_ump); } /* * Indirect block allocation dependencies. * * The same dependencies that exist for a direct block also exist when * a new block is allocated and pointed to by an entry in a block of * indirect pointers. The undo/redo states described above are also * used here. Because an indirect block contains many pointers that * may have dependencies, a second copy of the entire in-memory indirect * block is kept. The buffer cache copy is always completely up-to-date. * The second copy, which is used only as a source for disk writes, * contains only the safe pointers (i.e., those that have no remaining * update dependencies). The second copy is freed when all pointers * are safe. The cache is not allowed to replace indirect blocks with * pending update dependencies. If a buffer containing an indirect * block with dependencies is written, these routines will mark it * dirty again. It can only be successfully written once all the * dependencies are removed. The ffs_fsync routine in conjunction with * softdep_sync_metadata work together to get all the dependencies * removed so that a file can be successfully written to disk. Three * procedures are used when setting up indirect block pointer * dependencies. The division is necessary because of the organization * of the "balloc" routine and because of the distinction between file * pages and file metadata blocks. */ /* * Allocate a new allocindir structure. */ static struct allocindir * newallocindir(ip, ptrno, newblkno, oldblkno, lbn) struct inode *ip; /* inode for file being extended */ int ptrno; /* offset of pointer in indirect block */ ufs2_daddr_t newblkno; /* disk block number being added */ ufs2_daddr_t oldblkno; /* previous block number, 0 if none */ ufs_lbn_t lbn; { struct newblk *newblk; struct allocindir *aip; struct freefrag *freefrag; struct jnewblk *jnewblk; if (oldblkno) freefrag = newfreefrag(ip, oldblkno, ip->i_fs->fs_bsize, lbn); else freefrag = NULL; ACQUIRE_LOCK(ip->i_ump); if (newblk_lookup(UFSTOVFS(ip->i_ump), newblkno, 0, &newblk) == 0) panic("new_allocindir: lost block"); KASSERT(newblk->nb_list.wk_type == D_NEWBLK, ("newallocindir: newblk already initialized")); WORKITEM_REASSIGN(newblk, D_ALLOCINDIR); newblk->nb_freefrag = freefrag; aip = (struct allocindir *)newblk; aip->ai_offset = ptrno; aip->ai_oldblkno = oldblkno; aip->ai_lbn = lbn; if ((jnewblk = newblk->nb_jnewblk) != NULL) { jnewblk->jn_ino = ip->i_number; jnewblk->jn_lbn = lbn; add_to_journal(&jnewblk->jn_list); } if (freefrag && freefrag->ff_jdep != NULL && freefrag->ff_jdep->wk_type == D_JFREEFRAG) add_to_journal(freefrag->ff_jdep); return (aip); } /* * Called just before setting an indirect block pointer * to a newly allocated file page. */ void softdep_setup_allocindir_page(ip, lbn, bp, ptrno, newblkno, oldblkno, nbp) struct inode *ip; /* inode for file being extended */ ufs_lbn_t lbn; /* allocated block number within file */ struct buf *bp; /* buffer with indirect blk referencing page */ int ptrno; /* offset of pointer in indirect block */ ufs2_daddr_t newblkno; /* disk block number being added */ ufs2_daddr_t oldblkno; /* previous block number, 0 if none */ struct buf *nbp; /* buffer holding allocated page */ { struct inodedep *inodedep; struct freefrag *freefrag; struct allocindir *aip; struct pagedep *pagedep; struct mount *mp; mp = UFSTOVFS(ip->i_ump); KASSERT(MOUNTEDSOFTDEP(mp) != 0, ("softdep_setup_allocindir_page called on non-softdep filesystem")); KASSERT(lbn == nbp->b_lblkno, ("softdep_setup_allocindir_page: lbn %jd != lblkno %jd", lbn, bp->b_lblkno)); CTR4(KTR_SUJ, "softdep_setup_allocindir_page: ino %d blkno %jd oldblkno %jd " "lbn %jd", ip->i_number, newblkno, oldblkno, lbn); ASSERT_VOP_LOCKED(ITOV(ip), "softdep_setup_allocindir_page"); aip = newallocindir(ip, ptrno, newblkno, oldblkno, lbn); (void) inodedep_lookup(mp, ip->i_number, DEPALLOC, &inodedep); /* * If we are allocating a directory page, then we must * allocate an associated pagedep to track additions and * deletions. */ if ((ip->i_mode & IFMT) == IFDIR) pagedep_lookup(mp, nbp, ip->i_number, lbn, DEPALLOC, &pagedep); WORKLIST_INSERT(&nbp->b_dep, &aip->ai_block.nb_list); freefrag = setup_allocindir_phase2(bp, ip, inodedep, aip, lbn); FREE_LOCK(ip->i_ump); if (freefrag) handle_workitem_freefrag(freefrag); } /* * Called just before setting an indirect block pointer to a * newly allocated indirect block. */ void softdep_setup_allocindir_meta(nbp, ip, bp, ptrno, newblkno) struct buf *nbp; /* newly allocated indirect block */ struct inode *ip; /* inode for file being extended */ struct buf *bp; /* indirect block referencing allocated block */ int ptrno; /* offset of pointer in indirect block */ ufs2_daddr_t newblkno; /* disk block number being added */ { struct inodedep *inodedep; struct allocindir *aip; ufs_lbn_t lbn; KASSERT(MOUNTEDSOFTDEP(UFSTOVFS(ip->i_ump)) != 0, ("softdep_setup_allocindir_meta called on non-softdep filesystem")); CTR3(KTR_SUJ, "softdep_setup_allocindir_meta: ino %d blkno %jd ptrno %d", ip->i_number, newblkno, ptrno); lbn = nbp->b_lblkno; ASSERT_VOP_LOCKED(ITOV(ip), "softdep_setup_allocindir_meta"); aip = newallocindir(ip, ptrno, newblkno, 0, lbn); inodedep_lookup(UFSTOVFS(ip->i_ump), ip->i_number, DEPALLOC, &inodedep); WORKLIST_INSERT(&nbp->b_dep, &aip->ai_block.nb_list); if (setup_allocindir_phase2(bp, ip, inodedep, aip, lbn)) panic("softdep_setup_allocindir_meta: Block already existed"); FREE_LOCK(ip->i_ump); } static void indirdep_complete(indirdep) struct indirdep *indirdep; { struct allocindir *aip; LIST_REMOVE(indirdep, ir_next); indirdep->ir_state |= DEPCOMPLETE; while ((aip = LIST_FIRST(&indirdep->ir_completehd)) != NULL) { LIST_REMOVE(aip, ai_next); free_newblk(&aip->ai_block); } /* * If this indirdep is not attached to a buf it was simply waiting * on completion to clear completehd. free_indirdep() asserts * that nothing is dangling. */ if ((indirdep->ir_state & ONWORKLIST) == 0) free_indirdep(indirdep); } static struct indirdep * indirdep_lookup(mp, ip, bp) struct mount *mp; struct inode *ip; struct buf *bp; { struct indirdep *indirdep, *newindirdep; struct newblk *newblk; struct ufsmount *ump; struct worklist *wk; struct fs *fs; ufs2_daddr_t blkno; ump = VFSTOUFS(mp); LOCK_OWNED(ump); indirdep = NULL; newindirdep = NULL; fs = ip->i_fs; for (;;) { LIST_FOREACH(wk, &bp->b_dep, wk_list) { if (wk->wk_type != D_INDIRDEP) continue; indirdep = WK_INDIRDEP(wk); break; } /* Found on the buffer worklist, no new structure to free. */ if (indirdep != NULL && newindirdep == NULL) return (indirdep); if (indirdep != NULL && newindirdep != NULL) panic("indirdep_lookup: simultaneous create"); /* None found on the buffer and a new structure is ready. */ if (indirdep == NULL && newindirdep != NULL) break; /* None found and no new structure available. */ FREE_LOCK(ump); newindirdep = malloc(sizeof(struct indirdep), M_INDIRDEP, M_SOFTDEP_FLAGS); workitem_alloc(&newindirdep->ir_list, D_INDIRDEP, mp); newindirdep->ir_state = ATTACHED; if (ip->i_ump->um_fstype == UFS1) newindirdep->ir_state |= UFS1FMT; TAILQ_INIT(&newindirdep->ir_trunc); newindirdep->ir_saveddata = NULL; LIST_INIT(&newindirdep->ir_deplisthd); LIST_INIT(&newindirdep->ir_donehd); LIST_INIT(&newindirdep->ir_writehd); LIST_INIT(&newindirdep->ir_completehd); if (bp->b_blkno == bp->b_lblkno) { ufs_bmaparray(bp->b_vp, bp->b_lblkno, &blkno, bp, NULL, NULL); bp->b_blkno = blkno; } newindirdep->ir_freeblks = NULL; newindirdep->ir_savebp = getblk(ip->i_devvp, bp->b_blkno, bp->b_bcount, 0, 0, 0); newindirdep->ir_bp = bp; BUF_KERNPROC(newindirdep->ir_savebp); bcopy(bp->b_data, newindirdep->ir_savebp->b_data, bp->b_bcount); ACQUIRE_LOCK(ump); } indirdep = newindirdep; WORKLIST_INSERT(&bp->b_dep, &indirdep->ir_list); /* * If the block is not yet allocated we don't set DEPCOMPLETE so * that we don't free dependencies until the pointers are valid. * This could search b_dep for D_ALLOCDIRECT/D_ALLOCINDIR rather * than using the hash. */ if (newblk_lookup(mp, dbtofsb(fs, bp->b_blkno), 0, &newblk)) LIST_INSERT_HEAD(&newblk->nb_indirdeps, indirdep, ir_next); else indirdep->ir_state |= DEPCOMPLETE; return (indirdep); } /* * Called to finish the allocation of the "aip" allocated * by one of the two routines above. */ static struct freefrag * setup_allocindir_phase2(bp, ip, inodedep, aip, lbn) struct buf *bp; /* in-memory copy of the indirect block */ struct inode *ip; /* inode for file being extended */ struct inodedep *inodedep; /* Inodedep for ip */ struct allocindir *aip; /* allocindir allocated by the above routines */ ufs_lbn_t lbn; /* Logical block number for this block. */ { struct fs *fs; struct indirdep *indirdep; struct allocindir *oldaip; struct freefrag *freefrag; struct mount *mp; LOCK_OWNED(ip->i_ump); mp = UFSTOVFS(ip->i_ump); fs = ip->i_fs; if (bp->b_lblkno >= 0) panic("setup_allocindir_phase2: not indir blk"); KASSERT(aip->ai_offset >= 0 && aip->ai_offset < NINDIR(fs), ("setup_allocindir_phase2: Bad offset %d", aip->ai_offset)); indirdep = indirdep_lookup(mp, ip, bp); KASSERT(indirdep->ir_savebp != NULL, ("setup_allocindir_phase2 NULL ir_savebp")); aip->ai_indirdep = indirdep; /* * Check for an unwritten dependency for this indirect offset. If * there is, merge the old dependency into the new one. This happens * as a result of reallocblk only. */ freefrag = NULL; if (aip->ai_oldblkno != 0) { LIST_FOREACH(oldaip, &indirdep->ir_deplisthd, ai_next) { if (oldaip->ai_offset == aip->ai_offset) { freefrag = allocindir_merge(aip, oldaip); goto done; } } LIST_FOREACH(oldaip, &indirdep->ir_donehd, ai_next) { if (oldaip->ai_offset == aip->ai_offset) { freefrag = allocindir_merge(aip, oldaip); goto done; } } } done: LIST_INSERT_HEAD(&indirdep->ir_deplisthd, aip, ai_next); return (freefrag); } /* * Merge two allocindirs which refer to the same block. Move newblock * dependencies and setup the freefrags appropriately. */ static struct freefrag * allocindir_merge(aip, oldaip) struct allocindir *aip; struct allocindir *oldaip; { struct freefrag *freefrag; struct worklist *wk; if (oldaip->ai_newblkno != aip->ai_oldblkno) panic("allocindir_merge: blkno"); aip->ai_oldblkno = oldaip->ai_oldblkno; freefrag = aip->ai_freefrag; aip->ai_freefrag = oldaip->ai_freefrag; oldaip->ai_freefrag = NULL; KASSERT(freefrag != NULL, ("setup_allocindir_phase2: No freefrag")); /* * If we are tracking a new directory-block allocation, * move it from the old allocindir to the new allocindir. */ if ((wk = LIST_FIRST(&oldaip->ai_newdirblk)) != NULL) { WORKLIST_REMOVE(wk); if (!LIST_EMPTY(&oldaip->ai_newdirblk)) panic("allocindir_merge: extra newdirblk"); WORKLIST_INSERT(&aip->ai_newdirblk, wk); } /* * We can skip journaling for this freefrag and just complete * any pending journal work for the allocindir that is being * removed after the freefrag completes. */ if (freefrag->ff_jdep) cancel_jfreefrag(WK_JFREEFRAG(freefrag->ff_jdep)); LIST_REMOVE(oldaip, ai_next); freefrag->ff_jdep = (struct worklist *)cancel_newblk(&oldaip->ai_block, &freefrag->ff_list, &freefrag->ff_jwork); free_newblk(&oldaip->ai_block); return (freefrag); } static inline void setup_freedirect(freeblks, ip, i, needj) struct freeblks *freeblks; struct inode *ip; int i; int needj; { ufs2_daddr_t blkno; int frags; blkno = DIP(ip, i_db[i]); if (blkno == 0) return; DIP_SET(ip, i_db[i], 0); frags = sblksize(ip->i_fs, ip->i_size, i); frags = numfrags(ip->i_fs, frags); newfreework(ip->i_ump, freeblks, NULL, i, blkno, frags, 0, needj); } static inline void setup_freeext(freeblks, ip, i, needj) struct freeblks *freeblks; struct inode *ip; int i; int needj; { ufs2_daddr_t blkno; int frags; blkno = ip->i_din2->di_extb[i]; if (blkno == 0) return; ip->i_din2->di_extb[i] = 0; frags = sblksize(ip->i_fs, ip->i_din2->di_extsize, i); frags = numfrags(ip->i_fs, frags); newfreework(ip->i_ump, freeblks, NULL, -1 - i, blkno, frags, 0, needj); } static inline void setup_freeindir(freeblks, ip, i, lbn, needj) struct freeblks *freeblks; struct inode *ip; int i; ufs_lbn_t lbn; int needj; { ufs2_daddr_t blkno; blkno = DIP(ip, i_ib[i]); if (blkno == 0) return; DIP_SET(ip, i_ib[i], 0); newfreework(ip->i_ump, freeblks, NULL, lbn, blkno, ip->i_fs->fs_frag, 0, needj); } static inline struct freeblks * newfreeblks(mp, ip) struct mount *mp; struct inode *ip; { struct freeblks *freeblks; freeblks = malloc(sizeof(struct freeblks), M_FREEBLKS, M_SOFTDEP_FLAGS|M_ZERO); workitem_alloc(&freeblks->fb_list, D_FREEBLKS, mp); LIST_INIT(&freeblks->fb_jblkdephd); LIST_INIT(&freeblks->fb_jwork); freeblks->fb_ref = 0; freeblks->fb_cgwait = 0; freeblks->fb_state = ATTACHED; freeblks->fb_uid = ip->i_uid; freeblks->fb_inum = ip->i_number; freeblks->fb_vtype = ITOV(ip)->v_type; freeblks->fb_modrev = DIP(ip, i_modrev); freeblks->fb_devvp = ip->i_devvp; freeblks->fb_chkcnt = 0; freeblks->fb_len = 0; return (freeblks); } static void trunc_indirdep(indirdep, freeblks, bp, off) struct indirdep *indirdep; struct freeblks *freeblks; struct buf *bp; int off; { struct allocindir *aip, *aipn; /* * The first set of allocindirs won't be in savedbp. */ LIST_FOREACH_SAFE(aip, &indirdep->ir_deplisthd, ai_next, aipn) if (aip->ai_offset > off) cancel_allocindir(aip, bp, freeblks, 1); LIST_FOREACH_SAFE(aip, &indirdep->ir_donehd, ai_next, aipn) if (aip->ai_offset > off) cancel_allocindir(aip, bp, freeblks, 1); /* * These will exist in savedbp. */ LIST_FOREACH_SAFE(aip, &indirdep->ir_writehd, ai_next, aipn) if (aip->ai_offset > off) cancel_allocindir(aip, NULL, freeblks, 0); LIST_FOREACH_SAFE(aip, &indirdep->ir_completehd, ai_next, aipn) if (aip->ai_offset > off) cancel_allocindir(aip, NULL, freeblks, 0); } /* * Follow the chain of indirects down to lastlbn creating a freework * structure for each. This will be used to start indir_trunc() at * the right offset and create the journal records for the parrtial * truncation. A second step will handle the truncated dependencies. */ static int setup_trunc_indir(freeblks, ip, lbn, lastlbn, blkno) struct freeblks *freeblks; struct inode *ip; ufs_lbn_t lbn; ufs_lbn_t lastlbn; ufs2_daddr_t blkno; { struct indirdep *indirdep; struct indirdep *indirn; struct freework *freework; struct newblk *newblk; struct mount *mp; struct buf *bp; uint8_t *start; uint8_t *end; ufs_lbn_t lbnadd; int level; int error; int off; freework = NULL; if (blkno == 0) return (0); mp = freeblks->fb_list.wk_mp; bp = getblk(ITOV(ip), lbn, mp->mnt_stat.f_iosize, 0, 0, 0); if ((bp->b_flags & B_CACHE) == 0) { bp->b_blkno = blkptrtodb(VFSTOUFS(mp), blkno); bp->b_iocmd = BIO_READ; bp->b_flags &= ~B_INVAL; bp->b_ioflags &= ~BIO_ERROR; vfs_busy_pages(bp, 0); bp->b_iooffset = dbtob(bp->b_blkno); bstrategy(bp); curthread->td_ru.ru_inblock++; error = bufwait(bp); if (error) { brelse(bp); return (error); } } level = lbn_level(lbn); lbnadd = lbn_offset(ip->i_fs, level); /* * Compute the offset of the last block we want to keep. Store * in the freework the first block we want to completely free. */ off = (lastlbn - -(lbn + level)) / lbnadd; if (off + 1 == NINDIR(ip->i_fs)) goto nowork; freework = newfreework(ip->i_ump, freeblks, NULL, lbn, blkno, 0, off+1, 0); /* * Link the freework into the indirdep. This will prevent any new * allocations from proceeding until we are finished with the * truncate and the block is written. */ ACQUIRE_LOCK(ip->i_ump); indirdep = indirdep_lookup(mp, ip, bp); if (indirdep->ir_freeblks) panic("setup_trunc_indir: indirdep already truncated."); TAILQ_INSERT_TAIL(&indirdep->ir_trunc, freework, fw_next); freework->fw_indir = indirdep; /* * Cancel any allocindirs that will not make it to disk. * We have to do this for all copies of the indirdep that * live on this newblk. */ if ((indirdep->ir_state & DEPCOMPLETE) == 0) { newblk_lookup(mp, dbtofsb(ip->i_fs, bp->b_blkno), 0, &newblk); LIST_FOREACH(indirn, &newblk->nb_indirdeps, ir_next) trunc_indirdep(indirn, freeblks, bp, off); } else trunc_indirdep(indirdep, freeblks, bp, off); FREE_LOCK(ip->i_ump); /* * Creation is protected by the buf lock. The saveddata is only * needed if a full truncation follows a partial truncation but it * is difficult to allocate in that case so we fetch it anyway. */ if (indirdep->ir_saveddata == NULL) indirdep->ir_saveddata = malloc(bp->b_bcount, M_INDIRDEP, M_SOFTDEP_FLAGS); nowork: /* Fetch the blkno of the child and the zero start offset. */ if (ip->i_ump->um_fstype == UFS1) { blkno = ((ufs1_daddr_t *)bp->b_data)[off]; start = (uint8_t *)&((ufs1_daddr_t *)bp->b_data)[off+1]; } else { blkno = ((ufs2_daddr_t *)bp->b_data)[off]; start = (uint8_t *)&((ufs2_daddr_t *)bp->b_data)[off+1]; } if (freework) { /* Zero the truncated pointers. */ end = bp->b_data + bp->b_bcount; bzero(start, end - start); bdwrite(bp); } else bqrelse(bp); if (level == 0) return (0); lbn++; /* adjust level */ lbn -= (off * lbnadd); return setup_trunc_indir(freeblks, ip, lbn, lastlbn, blkno); } /* * Complete the partial truncation of an indirect block setup by * setup_trunc_indir(). This zeros the truncated pointers in the saved * copy and writes them to disk before the freeblks is allowed to complete. */ static void complete_trunc_indir(freework) struct freework *freework; { struct freework *fwn; struct indirdep *indirdep; struct ufsmount *ump; struct buf *bp; uintptr_t start; int count; ump = VFSTOUFS(freework->fw_list.wk_mp); LOCK_OWNED(ump); indirdep = freework->fw_indir; for (;;) { bp = indirdep->ir_bp; /* See if the block was discarded. */ if (bp == NULL) break; /* Inline part of getdirtybuf(). We dont want bremfree. */ if (BUF_LOCK(bp, LK_EXCLUSIVE | LK_NOWAIT, NULL) == 0) break; if (BUF_LOCK(bp, LK_EXCLUSIVE | LK_SLEEPFAIL | LK_INTERLOCK, LOCK_PTR(ump)) == 0) BUF_UNLOCK(bp); ACQUIRE_LOCK(ump); } freework->fw_state |= DEPCOMPLETE; TAILQ_REMOVE(&indirdep->ir_trunc, freework, fw_next); /* * Zero the pointers in the saved copy. */ if (indirdep->ir_state & UFS1FMT) start = sizeof(ufs1_daddr_t); else start = sizeof(ufs2_daddr_t); start *= freework->fw_start; count = indirdep->ir_savebp->b_bcount - start; start += (uintptr_t)indirdep->ir_savebp->b_data; bzero((char *)start, count); /* * We need to start the next truncation in the list if it has not * been started yet. */ fwn = TAILQ_FIRST(&indirdep->ir_trunc); if (fwn != NULL) { if (fwn->fw_freeblks == indirdep->ir_freeblks) TAILQ_REMOVE(&indirdep->ir_trunc, fwn, fw_next); if ((fwn->fw_state & ONWORKLIST) == 0) freework_enqueue(fwn); } /* * If bp is NULL the block was fully truncated, restore * the saved block list otherwise free it if it is no * longer needed. */ if (TAILQ_EMPTY(&indirdep->ir_trunc)) { if (bp == NULL) bcopy(indirdep->ir_saveddata, indirdep->ir_savebp->b_data, indirdep->ir_savebp->b_bcount); free(indirdep->ir_saveddata, M_INDIRDEP); indirdep->ir_saveddata = NULL; } /* * When bp is NULL there is a full truncation pending. We * must wait for this full truncation to be journaled before * we can release this freework because the disk pointers will * never be written as zero. */ if (bp == NULL) { if (LIST_EMPTY(&indirdep->ir_freeblks->fb_jblkdephd)) handle_written_freework(freework); else WORKLIST_INSERT(&indirdep->ir_freeblks->fb_freeworkhd, &freework->fw_list); } else { /* Complete when the real copy is written. */ WORKLIST_INSERT(&bp->b_dep, &freework->fw_list); BUF_UNLOCK(bp); } } /* * Calculate the number of blocks we are going to release where datablocks * is the current total and length is the new file size. */ static ufs2_daddr_t blkcount(fs, datablocks, length) struct fs *fs; ufs2_daddr_t datablocks; off_t length; { off_t totblks, numblks; totblks = 0; numblks = howmany(length, fs->fs_bsize); if (numblks <= NDADDR) { totblks = howmany(length, fs->fs_fsize); goto out; } totblks = blkstofrags(fs, numblks); numblks -= NDADDR; /* * Count all single, then double, then triple indirects required. * Subtracting one indirects worth of blocks for each pass * acknowledges one of each pointed to by the inode. */ for (;;) { totblks += blkstofrags(fs, howmany(numblks, NINDIR(fs))); numblks -= NINDIR(fs); if (numblks <= 0) break; numblks = howmany(numblks, NINDIR(fs)); } out: totblks = fsbtodb(fs, totblks); /* * Handle sparse files. We can't reclaim more blocks than the inode * references. We will correct it later in handle_complete_freeblks() * when we know the real count. */ if (totblks > datablocks) return (0); return (datablocks - totblks); } /* * Handle freeblocks for journaled softupdate filesystems. * * Contrary to normal softupdates, we must preserve the block pointers in * indirects until their subordinates are free. This is to avoid journaling * every block that is freed which may consume more space than the journal * itself. The recovery program will see the free block journals at the * base of the truncated area and traverse them to reclaim space. The * pointers in the inode may be cleared immediately after the journal * records are written because each direct and indirect pointer in the * inode is recorded in a journal. This permits full truncation to proceed * asynchronously. The write order is journal -> inode -> cgs -> indirects. * * The algorithm is as follows: * 1) Traverse the in-memory state and create journal entries to release * the relevant blocks and full indirect trees. * 2) Traverse the indirect block chain adding partial truncation freework * records to indirects in the path to lastlbn. The freework will * prevent new allocation dependencies from being satisfied in this * indirect until the truncation completes. * 3) Read and lock the inode block, performing an update with the new size * and pointers. This prevents truncated data from becoming valid on * disk through step 4. * 4) Reap unsatisfied dependencies that are beyond the truncated area, * eliminate journal work for those records that do not require it. * 5) Schedule the journal records to be written followed by the inode block. * 6) Allocate any necessary frags for the end of file. * 7) Zero any partially truncated blocks. * * From this truncation proceeds asynchronously using the freework and * indir_trunc machinery. The file will not be extended again into a * partially truncated indirect block until all work is completed but * the normal dependency mechanism ensures that it is rolled back/forward * as appropriate. Further truncation may occur without delay and is * serialized in indir_trunc(). */ void softdep_journal_freeblocks(ip, cred, length, flags) struct inode *ip; /* The inode whose length is to be reduced */ struct ucred *cred; off_t length; /* The new length for the file */ int flags; /* IO_EXT and/or IO_NORMAL */ { struct freeblks *freeblks, *fbn; struct worklist *wk, *wkn; struct inodedep *inodedep; struct jblkdep *jblkdep; struct allocdirect *adp, *adpn; struct ufsmount *ump; struct fs *fs; struct buf *bp; struct vnode *vp; struct mount *mp; ufs2_daddr_t extblocks, datablocks; ufs_lbn_t tmpval, lbn, lastlbn; int frags, lastoff, iboff, allocblock, needj, error, i; fs = ip->i_fs; ump = ip->i_ump; mp = UFSTOVFS(ump); KASSERT(MOUNTEDSOFTDEP(mp) != 0, ("softdep_journal_freeblocks called on non-softdep filesystem")); vp = ITOV(ip); needj = 1; iboff = -1; allocblock = 0; extblocks = 0; datablocks = 0; frags = 0; freeblks = newfreeblks(mp, ip); ACQUIRE_LOCK(ump); /* * If we're truncating a removed file that will never be written * we don't need to journal the block frees. The canceled journals * for the allocations will suffice. */ inodedep_lookup(mp, ip->i_number, DEPALLOC, &inodedep); if ((inodedep->id_state & (UNLINKED | DEPCOMPLETE)) == UNLINKED && length == 0) needj = 0; CTR3(KTR_SUJ, "softdep_journal_freeblks: ip %d length %ld needj %d", ip->i_number, length, needj); FREE_LOCK(ump); /* * Calculate the lbn that we are truncating to. This results in -1 * if we're truncating the 0 bytes. So it is the last lbn we want * to keep, not the first lbn we want to truncate. */ lastlbn = lblkno(fs, length + fs->fs_bsize - 1) - 1; lastoff = blkoff(fs, length); /* * Compute frags we are keeping in lastlbn. 0 means all. */ if (lastlbn >= 0 && lastlbn < NDADDR) { frags = fragroundup(fs, lastoff); /* adp offset of last valid allocdirect. */ iboff = lastlbn; } else if (lastlbn > 0) iboff = NDADDR; if (fs->fs_magic == FS_UFS2_MAGIC) extblocks = btodb(fragroundup(fs, ip->i_din2->di_extsize)); /* * Handle normal data blocks and indirects. This section saves * values used after the inode update to complete frag and indirect * truncation. */ if ((flags & IO_NORMAL) != 0) { /* * Handle truncation of whole direct and indirect blocks. */ for (i = iboff + 1; i < NDADDR; i++) setup_freedirect(freeblks, ip, i, needj); for (i = 0, tmpval = NINDIR(fs), lbn = NDADDR; i < NIADDR; i++, lbn += tmpval, tmpval *= NINDIR(fs)) { /* Release a whole indirect tree. */ if (lbn > lastlbn) { setup_freeindir(freeblks, ip, i, -lbn -i, needj); continue; } iboff = i + NDADDR; /* * Traverse partially truncated indirect tree. */ if (lbn <= lastlbn && lbn + tmpval - 1 > lastlbn) setup_trunc_indir(freeblks, ip, -lbn - i, lastlbn, DIP(ip, i_ib[i])); } /* * Handle partial truncation to a frag boundary. */ if (frags) { ufs2_daddr_t blkno; long oldfrags; oldfrags = blksize(fs, ip, lastlbn); blkno = DIP(ip, i_db[lastlbn]); if (blkno && oldfrags != frags) { oldfrags -= frags; oldfrags = numfrags(ip->i_fs, oldfrags); blkno += numfrags(ip->i_fs, frags); newfreework(ump, freeblks, NULL, lastlbn, blkno, oldfrags, 0, needj); if (needj) adjust_newfreework(freeblks, numfrags(ip->i_fs, frags)); } else if (blkno == 0) allocblock = 1; } /* * Add a journal record for partial truncate if we are * handling indirect blocks. Non-indirects need no extra * journaling. */ if (length != 0 && lastlbn >= NDADDR) { ip->i_flag |= IN_TRUNCATED; newjtrunc(freeblks, length, 0); } ip->i_size = length; DIP_SET(ip, i_size, ip->i_size); datablocks = DIP(ip, i_blocks) - extblocks; if (length != 0) datablocks = blkcount(ip->i_fs, datablocks, length); freeblks->fb_len = length; } if ((flags & IO_EXT) != 0) { for (i = 0; i < NXADDR; i++) setup_freeext(freeblks, ip, i, needj); ip->i_din2->di_extsize = 0; datablocks += extblocks; } #ifdef QUOTA /* Reference the quotas in case the block count is wrong in the end. */ quotaref(vp, freeblks->fb_quota); (void) chkdq(ip, -datablocks, NOCRED, 0); #endif freeblks->fb_chkcnt = -datablocks; UFS_LOCK(ump); fs->fs_pendingblocks += datablocks; UFS_UNLOCK(ump); DIP_SET(ip, i_blocks, DIP(ip, i_blocks) - datablocks); /* * Handle truncation of incomplete alloc direct dependencies. We * hold the inode block locked to prevent incomplete dependencies * from reaching the disk while we are eliminating those that * have been truncated. This is a partially inlined ffs_update(). */ ufs_itimes(vp); ip->i_flag &= ~(IN_LAZYACCESS | IN_LAZYMOD | IN_MODIFIED); error = bread(ip->i_devvp, fsbtodb(fs, ino_to_fsba(fs, ip->i_number)), (int)fs->fs_bsize, cred, &bp); if (error) { brelse(bp); softdep_error("softdep_journal_freeblocks", error); return; } if (bp->b_bufsize == fs->fs_bsize) bp->b_flags |= B_CLUSTEROK; softdep_update_inodeblock(ip, bp, 0); if (ump->um_fstype == UFS1) *((struct ufs1_dinode *)bp->b_data + ino_to_fsbo(fs, ip->i_number)) = *ip->i_din1; else *((struct ufs2_dinode *)bp->b_data + ino_to_fsbo(fs, ip->i_number)) = *ip->i_din2; ACQUIRE_LOCK(ump); (void) inodedep_lookup(mp, ip->i_number, DEPALLOC, &inodedep); if ((inodedep->id_state & IOSTARTED) != 0) panic("softdep_setup_freeblocks: inode busy"); /* * Add the freeblks structure to the list of operations that * must await the zero'ed inode being written to disk. If we * still have a bitmap dependency (needj), then the inode * has never been written to disk, so we can process the * freeblks below once we have deleted the dependencies. */ if (needj) WORKLIST_INSERT(&bp->b_dep, &freeblks->fb_list); else freeblks->fb_state |= COMPLETE; if ((flags & IO_NORMAL) != 0) { TAILQ_FOREACH_SAFE(adp, &inodedep->id_inoupdt, ad_next, adpn) { if (adp->ad_offset > iboff) cancel_allocdirect(&inodedep->id_inoupdt, adp, freeblks); /* * Truncate the allocdirect. We could eliminate * or modify journal records as well. */ else if (adp->ad_offset == iboff && frags) adp->ad_newsize = frags; } } if ((flags & IO_EXT) != 0) while ((adp = TAILQ_FIRST(&inodedep->id_extupdt)) != 0) cancel_allocdirect(&inodedep->id_extupdt, adp, freeblks); /* * Scan the bufwait list for newblock dependencies that will never * make it to disk. */ LIST_FOREACH_SAFE(wk, &inodedep->id_bufwait, wk_list, wkn) { if (wk->wk_type != D_ALLOCDIRECT) continue; adp = WK_ALLOCDIRECT(wk); if (((flags & IO_NORMAL) != 0 && (adp->ad_offset > iboff)) || ((flags & IO_EXT) != 0 && (adp->ad_state & EXTDATA))) { cancel_jfreeblk(freeblks, adp->ad_newblkno); cancel_newblk(WK_NEWBLK(wk), NULL, &freeblks->fb_jwork); WORKLIST_INSERT(&freeblks->fb_freeworkhd, wk); } } /* * Add journal work. */ LIST_FOREACH(jblkdep, &freeblks->fb_jblkdephd, jb_deps) add_to_journal(&jblkdep->jb_list); FREE_LOCK(ump); bdwrite(bp); /* * Truncate dependency structures beyond length. */ trunc_dependencies(ip, freeblks, lastlbn, frags, flags); /* * This is only set when we need to allocate a fragment because * none existed at the end of a frag-sized file. It handles only * allocating a new, zero filled block. */ if (allocblock) { ip->i_size = length - lastoff; DIP_SET(ip, i_size, ip->i_size); error = UFS_BALLOC(vp, length - 1, 1, cred, BA_CLRBUF, &bp); if (error != 0) { softdep_error("softdep_journal_freeblks", error); return; } ip->i_size = length; DIP_SET(ip, i_size, length); ip->i_flag |= IN_CHANGE | IN_UPDATE; allocbuf(bp, frags); ffs_update(vp, 0); bawrite(bp); } else if (lastoff != 0 && vp->v_type != VDIR) { int size; /* * Zero the end of a truncated frag or block. */ size = sblksize(fs, length, lastlbn); error = bread(vp, lastlbn, size, cred, &bp); if (error) { softdep_error("softdep_journal_freeblks", error); return; } bzero((char *)bp->b_data + lastoff, size - lastoff); bawrite(bp); } ACQUIRE_LOCK(ump); inodedep_lookup(mp, ip->i_number, DEPALLOC, &inodedep); TAILQ_INSERT_TAIL(&inodedep->id_freeblklst, freeblks, fb_next); freeblks->fb_state |= DEPCOMPLETE | ONDEPLIST; /* * We zero earlier truncations so they don't erroneously * update i_blocks. */ if (freeblks->fb_len == 0 && (flags & IO_NORMAL) != 0) TAILQ_FOREACH(fbn, &inodedep->id_freeblklst, fb_next) fbn->fb_len = 0; if ((freeblks->fb_state & ALLCOMPLETE) == ALLCOMPLETE && LIST_EMPTY(&freeblks->fb_jblkdephd)) freeblks->fb_state |= INPROGRESS; else freeblks = NULL; FREE_LOCK(ump); if (freeblks) handle_workitem_freeblocks(freeblks, 0); trunc_pages(ip, length, extblocks, flags); } /* * Flush a JOP_SYNC to the journal. */ void softdep_journal_fsync(ip) struct inode *ip; { struct jfsync *jfsync; KASSERT(MOUNTEDSOFTDEP(UFSTOVFS(ip->i_ump)) != 0, ("softdep_journal_fsync called on non-softdep filesystem")); if ((ip->i_flag & IN_TRUNCATED) == 0) return; ip->i_flag &= ~IN_TRUNCATED; jfsync = malloc(sizeof(*jfsync), M_JFSYNC, M_SOFTDEP_FLAGS | M_ZERO); workitem_alloc(&jfsync->jfs_list, D_JFSYNC, UFSTOVFS(ip->i_ump)); jfsync->jfs_size = ip->i_size; jfsync->jfs_ino = ip->i_number; ACQUIRE_LOCK(ip->i_ump); add_to_journal(&jfsync->jfs_list); jwait(&jfsync->jfs_list, MNT_WAIT); FREE_LOCK(ip->i_ump); } /* * Block de-allocation dependencies. * * When blocks are de-allocated, the on-disk pointers must be nullified before * the blocks are made available for use by other files. (The true * requirement is that old pointers must be nullified before new on-disk * pointers are set. We chose this slightly more stringent requirement to * reduce complexity.) Our implementation handles this dependency by updating * the inode (or indirect block) appropriately but delaying the actual block * de-allocation (i.e., freemap and free space count manipulation) until * after the updated versions reach stable storage. After the disk is * updated, the blocks can be safely de-allocated whenever it is convenient. * This implementation handles only the common case of reducing a file's * length to zero. Other cases are handled by the conventional synchronous * write approach. * * The ffs implementation with which we worked double-checks * the state of the block pointers and file size as it reduces * a file's length. Some of this code is replicated here in our * soft updates implementation. The freeblks->fb_chkcnt field is * used to transfer a part of this information to the procedure * that eventually de-allocates the blocks. * * This routine should be called from the routine that shortens * a file's length, before the inode's size or block pointers * are modified. It will save the block pointer information for * later release and zero the inode so that the calling routine * can release it. */ void softdep_setup_freeblocks(ip, length, flags) struct inode *ip; /* The inode whose length is to be reduced */ off_t length; /* The new length for the file */ int flags; /* IO_EXT and/or IO_NORMAL */ { struct ufs1_dinode *dp1; struct ufs2_dinode *dp2; struct freeblks *freeblks; struct inodedep *inodedep; struct allocdirect *adp; struct ufsmount *ump; struct buf *bp; struct fs *fs; ufs2_daddr_t extblocks, datablocks; struct mount *mp; int i, delay, error; ufs_lbn_t tmpval; ufs_lbn_t lbn; ump = ip->i_ump; mp = UFSTOVFS(ump); KASSERT(MOUNTEDSOFTDEP(mp) != 0, ("softdep_setup_freeblocks called on non-softdep filesystem")); CTR2(KTR_SUJ, "softdep_setup_freeblks: ip %d length %ld", ip->i_number, length); KASSERT(length == 0, ("softdep_setup_freeblocks: non-zero length")); fs = ip->i_fs; freeblks = newfreeblks(mp, ip); extblocks = 0; datablocks = 0; if (fs->fs_magic == FS_UFS2_MAGIC) extblocks = btodb(fragroundup(fs, ip->i_din2->di_extsize)); if ((flags & IO_NORMAL) != 0) { for (i = 0; i < NDADDR; i++) setup_freedirect(freeblks, ip, i, 0); for (i = 0, tmpval = NINDIR(fs), lbn = NDADDR; i < NIADDR; i++, lbn += tmpval, tmpval *= NINDIR(fs)) setup_freeindir(freeblks, ip, i, -lbn -i, 0); ip->i_size = 0; DIP_SET(ip, i_size, 0); datablocks = DIP(ip, i_blocks) - extblocks; } if ((flags & IO_EXT) != 0) { for (i = 0; i < NXADDR; i++) setup_freeext(freeblks, ip, i, 0); ip->i_din2->di_extsize = 0; datablocks += extblocks; } #ifdef QUOTA /* Reference the quotas in case the block count is wrong in the end. */ quotaref(ITOV(ip), freeblks->fb_quota); (void) chkdq(ip, -datablocks, NOCRED, 0); #endif freeblks->fb_chkcnt = -datablocks; UFS_LOCK(ump); fs->fs_pendingblocks += datablocks; UFS_UNLOCK(ump); DIP_SET(ip, i_blocks, DIP(ip, i_blocks) - datablocks); /* * Push the zero'ed inode to to its disk buffer so that we are free * to delete its dependencies below. Once the dependencies are gone * the buffer can be safely released. */ if ((error = bread(ip->i_devvp, fsbtodb(fs, ino_to_fsba(fs, ip->i_number)), (int)fs->fs_bsize, NOCRED, &bp)) != 0) { brelse(bp); softdep_error("softdep_setup_freeblocks", error); } if (ump->um_fstype == UFS1) { dp1 = ((struct ufs1_dinode *)bp->b_data + ino_to_fsbo(fs, ip->i_number)); ip->i_din1->di_freelink = dp1->di_freelink; *dp1 = *ip->i_din1; } else { dp2 = ((struct ufs2_dinode *)bp->b_data + ino_to_fsbo(fs, ip->i_number)); ip->i_din2->di_freelink = dp2->di_freelink; *dp2 = *ip->i_din2; } /* * Find and eliminate any inode dependencies. */ ACQUIRE_LOCK(ump); (void) inodedep_lookup(mp, ip->i_number, DEPALLOC, &inodedep); if ((inodedep->id_state & IOSTARTED) != 0) panic("softdep_setup_freeblocks: inode busy"); /* * Add the freeblks structure to the list of operations that * must await the zero'ed inode being written to disk. If we * still have a bitmap dependency (delay == 0), then the inode * has never been written to disk, so we can process the * freeblks below once we have deleted the dependencies. */ delay = (inodedep->id_state & DEPCOMPLETE); if (delay) WORKLIST_INSERT(&bp->b_dep, &freeblks->fb_list); else freeblks->fb_state |= COMPLETE; /* * Because the file length has been truncated to zero, any * pending block allocation dependency structures associated * with this inode are obsolete and can simply be de-allocated. * We must first merge the two dependency lists to get rid of * any duplicate freefrag structures, then purge the merged list. * If we still have a bitmap dependency, then the inode has never * been written to disk, so we can free any fragments without delay. */ if (flags & IO_NORMAL) { merge_inode_lists(&inodedep->id_newinoupdt, &inodedep->id_inoupdt); while ((adp = TAILQ_FIRST(&inodedep->id_inoupdt)) != 0) cancel_allocdirect(&inodedep->id_inoupdt, adp, freeblks); } if (flags & IO_EXT) { merge_inode_lists(&inodedep->id_newextupdt, &inodedep->id_extupdt); while ((adp = TAILQ_FIRST(&inodedep->id_extupdt)) != 0) cancel_allocdirect(&inodedep->id_extupdt, adp, freeblks); } FREE_LOCK(ump); bdwrite(bp); trunc_dependencies(ip, freeblks, -1, 0, flags); ACQUIRE_LOCK(ump); if (inodedep_lookup(mp, ip->i_number, 0, &inodedep) != 0) (void) free_inodedep(inodedep); freeblks->fb_state |= DEPCOMPLETE; /* * If the inode with zeroed block pointers is now on disk * we can start freeing blocks. */ if ((freeblks->fb_state & ALLCOMPLETE) == ALLCOMPLETE) freeblks->fb_state |= INPROGRESS; else freeblks = NULL; FREE_LOCK(ump); if (freeblks) handle_workitem_freeblocks(freeblks, 0); trunc_pages(ip, length, extblocks, flags); } /* * Eliminate pages from the page cache that back parts of this inode and * adjust the vnode pager's idea of our size. This prevents stale data * from hanging around in the page cache. */ static void trunc_pages(ip, length, extblocks, flags) struct inode *ip; off_t length; ufs2_daddr_t extblocks; int flags; { struct vnode *vp; struct fs *fs; ufs_lbn_t lbn; off_t end, extend; vp = ITOV(ip); fs = ip->i_fs; extend = OFF_TO_IDX(lblktosize(fs, -extblocks)); if ((flags & IO_EXT) != 0) vn_pages_remove(vp, extend, 0); if ((flags & IO_NORMAL) == 0) return; BO_LOCK(&vp->v_bufobj); drain_output(vp); BO_UNLOCK(&vp->v_bufobj); /* * The vnode pager eliminates file pages we eliminate indirects * below. */ vnode_pager_setsize(vp, length); /* * Calculate the end based on the last indirect we want to keep. If * the block extends into indirects we can just use the negative of * its lbn. Doubles and triples exist at lower numbers so we must * be careful not to remove those, if they exist. double and triple * indirect lbns do not overlap with others so it is not important * to verify how many levels are required. */ lbn = lblkno(fs, length); if (lbn >= NDADDR) { /* Calculate the virtual lbn of the triple indirect. */ lbn = -lbn - (NIADDR - 1); end = OFF_TO_IDX(lblktosize(fs, lbn)); } else end = extend; vn_pages_remove(vp, OFF_TO_IDX(OFF_MAX), end); } /* * See if the buf bp is in the range eliminated by truncation. */ static int trunc_check_buf(bp, blkoffp, lastlbn, lastoff, flags) struct buf *bp; int *blkoffp; ufs_lbn_t lastlbn; int lastoff; int flags; { ufs_lbn_t lbn; *blkoffp = 0; /* Only match ext/normal blocks as appropriate. */ if (((flags & IO_EXT) == 0 && (bp->b_xflags & BX_ALTDATA)) || ((flags & IO_NORMAL) == 0 && (bp->b_xflags & BX_ALTDATA) == 0)) return (0); /* ALTDATA is always a full truncation. */ if ((bp->b_xflags & BX_ALTDATA) != 0) return (1); /* -1 is full truncation. */ if (lastlbn == -1) return (1); /* * If this is a partial truncate we only want those * blocks and indirect blocks that cover the range * we're after. */ lbn = bp->b_lblkno; if (lbn < 0) lbn = -(lbn + lbn_level(lbn)); if (lbn < lastlbn) return (0); /* Here we only truncate lblkno if it's partial. */ if (lbn == lastlbn) { if (lastoff == 0) return (0); *blkoffp = lastoff; } return (1); } /* * Eliminate any dependencies that exist in memory beyond lblkno:off */ static void trunc_dependencies(ip, freeblks, lastlbn, lastoff, flags) struct inode *ip; struct freeblks *freeblks; ufs_lbn_t lastlbn; int lastoff; int flags; { struct bufobj *bo; struct vnode *vp; struct buf *bp; int blkoff; /* * We must wait for any I/O in progress to finish so that * all potential buffers on the dirty list will be visible. * Once they are all there, walk the list and get rid of * any dependencies. */ vp = ITOV(ip); bo = &vp->v_bufobj; BO_LOCK(bo); drain_output(vp); TAILQ_FOREACH(bp, &bo->bo_dirty.bv_hd, b_bobufs) bp->b_vflags &= ~BV_SCANNED; restart: TAILQ_FOREACH(bp, &bo->bo_dirty.bv_hd, b_bobufs) { if (bp->b_vflags & BV_SCANNED) continue; if (!trunc_check_buf(bp, &blkoff, lastlbn, lastoff, flags)) { bp->b_vflags |= BV_SCANNED; continue; } KASSERT(bp->b_bufobj == bo, ("Wrong object in buffer")); if ((bp = getdirtybuf(bp, BO_LOCKPTR(bo), MNT_WAIT)) == NULL) goto restart; BO_UNLOCK(bo); if (deallocate_dependencies(bp, freeblks, blkoff)) bqrelse(bp); else brelse(bp); BO_LOCK(bo); goto restart; } /* * Now do the work of vtruncbuf while also matching indirect blocks. */ TAILQ_FOREACH(bp, &bo->bo_clean.bv_hd, b_bobufs) bp->b_vflags &= ~BV_SCANNED; cleanrestart: TAILQ_FOREACH(bp, &bo->bo_clean.bv_hd, b_bobufs) { if (bp->b_vflags & BV_SCANNED) continue; if (!trunc_check_buf(bp, &blkoff, lastlbn, lastoff, flags)) { bp->b_vflags |= BV_SCANNED; continue; } if (BUF_LOCK(bp, LK_EXCLUSIVE | LK_SLEEPFAIL | LK_INTERLOCK, BO_LOCKPTR(bo)) == ENOLCK) { BO_LOCK(bo); goto cleanrestart; } bp->b_vflags |= BV_SCANNED; bremfree(bp); if (blkoff != 0) { allocbuf(bp, blkoff); bqrelse(bp); } else { bp->b_flags |= B_INVAL | B_NOCACHE | B_RELBUF; brelse(bp); } BO_LOCK(bo); goto cleanrestart; } drain_output(vp); BO_UNLOCK(bo); } static int cancel_pagedep(pagedep, freeblks, blkoff) struct pagedep *pagedep; struct freeblks *freeblks; int blkoff; { struct jremref *jremref; struct jmvref *jmvref; struct dirrem *dirrem, *tmp; int i; /* * Copy any directory remove dependencies to the list * to be processed after the freeblks proceeds. If * directory entry never made it to disk they * can be dumped directly onto the work list. */ LIST_FOREACH_SAFE(dirrem, &pagedep->pd_dirremhd, dm_next, tmp) { /* Skip this directory removal if it is intended to remain. */ if (dirrem->dm_offset < blkoff) continue; /* * If there are any dirrems we wait for the journal write * to complete and then restart the buf scan as the lock * has been dropped. */ while ((jremref = LIST_FIRST(&dirrem->dm_jremrefhd)) != NULL) { jwait(&jremref->jr_list, MNT_WAIT); return (ERESTART); } LIST_REMOVE(dirrem, dm_next); dirrem->dm_dirinum = pagedep->pd_ino; WORKLIST_INSERT(&freeblks->fb_freeworkhd, &dirrem->dm_list); } while ((jmvref = LIST_FIRST(&pagedep->pd_jmvrefhd)) != NULL) { jwait(&jmvref->jm_list, MNT_WAIT); return (ERESTART); } /* * When we're partially truncating a pagedep we just want to flush * journal entries and return. There can not be any adds in the * truncated portion of the directory and newblk must remain if * part of the block remains. */ if (blkoff != 0) { struct diradd *dap; LIST_FOREACH(dap, &pagedep->pd_pendinghd, da_pdlist) if (dap->da_offset > blkoff) panic("cancel_pagedep: diradd %p off %d > %d", dap, dap->da_offset, blkoff); for (i = 0; i < DAHASHSZ; i++) LIST_FOREACH(dap, &pagedep->pd_diraddhd[i], da_pdlist) if (dap->da_offset > blkoff) panic("cancel_pagedep: diradd %p off %d > %d", dap, dap->da_offset, blkoff); return (0); } /* * There should be no directory add dependencies present * as the directory could not be truncated until all * children were removed. */ KASSERT(LIST_FIRST(&pagedep->pd_pendinghd) == NULL, ("deallocate_dependencies: pendinghd != NULL")); for (i = 0; i < DAHASHSZ; i++) KASSERT(LIST_FIRST(&pagedep->pd_diraddhd[i]) == NULL, ("deallocate_dependencies: diraddhd != NULL")); if ((pagedep->pd_state & NEWBLOCK) != 0) free_newdirblk(pagedep->pd_newdirblk); if (free_pagedep(pagedep) == 0) panic("Failed to free pagedep %p", pagedep); return (0); } /* * Reclaim any dependency structures from a buffer that is about to * be reallocated to a new vnode. The buffer must be locked, thus, * no I/O completion operations can occur while we are manipulating * its associated dependencies. The mutex is held so that other I/O's * associated with related dependencies do not occur. */ static int deallocate_dependencies(bp, freeblks, off) struct buf *bp; struct freeblks *freeblks; int off; { struct indirdep *indirdep; struct pagedep *pagedep; - struct allocdirect *adp; struct worklist *wk, *wkn; struct ufsmount *ump; if ((wk = LIST_FIRST(&bp->b_dep)) == NULL) goto done; ump = VFSTOUFS(wk->wk_mp); ACQUIRE_LOCK(ump); LIST_FOREACH_SAFE(wk, &bp->b_dep, wk_list, wkn) { switch (wk->wk_type) { case D_INDIRDEP: indirdep = WK_INDIRDEP(wk); if (bp->b_lblkno >= 0 || bp->b_blkno != indirdep->ir_savebp->b_lblkno) panic("deallocate_dependencies: not indir"); cancel_indirdep(indirdep, bp, freeblks); continue; case D_PAGEDEP: pagedep = WK_PAGEDEP(wk); if (cancel_pagedep(pagedep, freeblks, off)) { FREE_LOCK(ump); return (ERESTART); } continue; case D_ALLOCINDIR: /* * Simply remove the allocindir, we'll find it via * the indirdep where we can clear pointers if * needed. */ WORKLIST_REMOVE(wk); continue; case D_FREEWORK: /* * A truncation is waiting for the zero'd pointers * to be written. It can be freed when the freeblks * is journaled. */ WORKLIST_REMOVE(wk); wk->wk_state |= ONDEPLIST; WORKLIST_INSERT(&freeblks->fb_freeworkhd, wk); break; case D_ALLOCDIRECT: - adp = WK_ALLOCDIRECT(wk); if (off != 0) continue; /* FALLTHROUGH */ default: panic("deallocate_dependencies: Unexpected type %s", TYPENAME(wk->wk_type)); /* NOTREACHED */ } } FREE_LOCK(ump); done: /* * Don't throw away this buf, we were partially truncating and * some deps may always remain. */ if (off) { allocbuf(bp, off); bp->b_vflags |= BV_SCANNED; return (EBUSY); } bp->b_flags |= B_INVAL | B_NOCACHE; return (0); } /* * An allocdirect is being canceled due to a truncate. We must make sure * the journal entry is released in concert with the blkfree that releases * the storage. Completed journal entries must not be released until the * space is no longer pointed to by the inode or in the bitmap. */ static void cancel_allocdirect(adphead, adp, freeblks) struct allocdirectlst *adphead; struct allocdirect *adp; struct freeblks *freeblks; { struct freework *freework; struct newblk *newblk; struct worklist *wk; TAILQ_REMOVE(adphead, adp, ad_next); newblk = (struct newblk *)adp; freework = NULL; /* * Find the correct freework structure. */ LIST_FOREACH(wk, &freeblks->fb_freeworkhd, wk_list) { if (wk->wk_type != D_FREEWORK) continue; freework = WK_FREEWORK(wk); if (freework->fw_blkno == newblk->nb_newblkno) break; } if (freework == NULL) panic("cancel_allocdirect: Freework not found"); /* * If a newblk exists at all we still have the journal entry that * initiated the allocation so we do not need to journal the free. */ cancel_jfreeblk(freeblks, freework->fw_blkno); /* * If the journal hasn't been written the jnewblk must be passed * to the call to ffs_blkfree that reclaims the space. We accomplish * this by linking the journal dependency into the freework to be * freed when freework_freeblock() is called. If the journal has * been written we can simply reclaim the journal space when the * freeblks work is complete. */ freework->fw_jnewblk = cancel_newblk(newblk, &freework->fw_list, &freeblks->fb_jwork); WORKLIST_INSERT(&freeblks->fb_freeworkhd, &newblk->nb_list); } /* * Cancel a new block allocation. May be an indirect or direct block. We * remove it from various lists and return any journal record that needs to * be resolved by the caller. * * A special consideration is made for indirects which were never pointed * at on disk and will never be found once this block is released. */ static struct jnewblk * cancel_newblk(newblk, wk, wkhd) struct newblk *newblk; struct worklist *wk; struct workhead *wkhd; { struct jnewblk *jnewblk; CTR1(KTR_SUJ, "cancel_newblk: blkno %jd", newblk->nb_newblkno); newblk->nb_state |= GOINGAWAY; /* * Previously we traversed the completedhd on each indirdep * attached to this newblk to cancel them and gather journal * work. Since we need only the oldest journal segment and * the lowest point on the tree will always have the oldest * journal segment we are free to release the segments * of any subordinates and may leave the indirdep list to * indirdep_complete() when this newblk is freed. */ if (newblk->nb_state & ONDEPLIST) { newblk->nb_state &= ~ONDEPLIST; LIST_REMOVE(newblk, nb_deps); } if (newblk->nb_state & ONWORKLIST) WORKLIST_REMOVE(&newblk->nb_list); /* * If the journal entry hasn't been written we save a pointer to * the dependency that frees it until it is written or the * superseding operation completes. */ jnewblk = newblk->nb_jnewblk; if (jnewblk != NULL && wk != NULL) { newblk->nb_jnewblk = NULL; jnewblk->jn_dep = wk; } if (!LIST_EMPTY(&newblk->nb_jwork)) jwork_move(wkhd, &newblk->nb_jwork); /* * When truncating we must free the newdirblk early to remove * the pagedep from the hash before returning. */ if ((wk = LIST_FIRST(&newblk->nb_newdirblk)) != NULL) free_newdirblk(WK_NEWDIRBLK(wk)); if (!LIST_EMPTY(&newblk->nb_newdirblk)) panic("cancel_newblk: extra newdirblk"); return (jnewblk); } /* * Schedule the freefrag associated with a newblk to be released once * the pointers are written and the previous block is no longer needed. */ static void newblk_freefrag(newblk) struct newblk *newblk; { struct freefrag *freefrag; if (newblk->nb_freefrag == NULL) return; freefrag = newblk->nb_freefrag; newblk->nb_freefrag = NULL; freefrag->ff_state |= COMPLETE; if ((freefrag->ff_state & ALLCOMPLETE) == ALLCOMPLETE) add_to_worklist(&freefrag->ff_list, 0); } /* * Free a newblk. Generate a new freefrag work request if appropriate. * This must be called after the inode pointer and any direct block pointers * are valid or fully removed via truncate or frag extension. */ static void free_newblk(newblk) struct newblk *newblk; { struct indirdep *indirdep; struct worklist *wk; KASSERT(newblk->nb_jnewblk == NULL, ("free_newblk: jnewblk %p still attached", newblk->nb_jnewblk)); KASSERT(newblk->nb_list.wk_type != D_NEWBLK, ("free_newblk: unclaimed newblk")); LOCK_OWNED(VFSTOUFS(newblk->nb_list.wk_mp)); newblk_freefrag(newblk); if (newblk->nb_state & ONDEPLIST) LIST_REMOVE(newblk, nb_deps); if (newblk->nb_state & ONWORKLIST) WORKLIST_REMOVE(&newblk->nb_list); LIST_REMOVE(newblk, nb_hash); if ((wk = LIST_FIRST(&newblk->nb_newdirblk)) != NULL) free_newdirblk(WK_NEWDIRBLK(wk)); if (!LIST_EMPTY(&newblk->nb_newdirblk)) panic("free_newblk: extra newdirblk"); while ((indirdep = LIST_FIRST(&newblk->nb_indirdeps)) != NULL) indirdep_complete(indirdep); handle_jwork(&newblk->nb_jwork); WORKITEM_FREE(newblk, D_NEWBLK); } /* * Free a newdirblk. Clear the NEWBLOCK flag on its associated pagedep. * This routine must be called with splbio interrupts blocked. */ static void free_newdirblk(newdirblk) struct newdirblk *newdirblk; { struct pagedep *pagedep; struct diradd *dap; struct worklist *wk; LOCK_OWNED(VFSTOUFS(newdirblk->db_list.wk_mp)); WORKLIST_REMOVE(&newdirblk->db_list); /* * If the pagedep is still linked onto the directory buffer * dependency chain, then some of the entries on the * pd_pendinghd list may not be committed to disk yet. In * this case, we will simply clear the NEWBLOCK flag and * let the pd_pendinghd list be processed when the pagedep * is next written. If the pagedep is no longer on the buffer * dependency chain, then all the entries on the pd_pending * list are committed to disk and we can free them here. */ pagedep = newdirblk->db_pagedep; pagedep->pd_state &= ~NEWBLOCK; if ((pagedep->pd_state & ONWORKLIST) == 0) { while ((dap = LIST_FIRST(&pagedep->pd_pendinghd)) != NULL) free_diradd(dap, NULL); /* * If no dependencies remain, the pagedep will be freed. */ free_pagedep(pagedep); } /* Should only ever be one item in the list. */ while ((wk = LIST_FIRST(&newdirblk->db_mkdir)) != NULL) { WORKLIST_REMOVE(wk); handle_written_mkdir(WK_MKDIR(wk), MKDIR_BODY); } WORKITEM_FREE(newdirblk, D_NEWDIRBLK); } /* * Prepare an inode to be freed. The actual free operation is not * done until the zero'ed inode has been written to disk. */ void softdep_freefile(pvp, ino, mode) struct vnode *pvp; ino_t ino; int mode; { struct inode *ip = VTOI(pvp); struct inodedep *inodedep; struct freefile *freefile; struct freeblks *freeblks; struct ufsmount *ump; ump = ip->i_ump; KASSERT(MOUNTEDSOFTDEP(UFSTOVFS(ump)) != 0, ("softdep_freefile called on non-softdep filesystem")); /* * This sets up the inode de-allocation dependency. */ freefile = malloc(sizeof(struct freefile), M_FREEFILE, M_SOFTDEP_FLAGS); workitem_alloc(&freefile->fx_list, D_FREEFILE, pvp->v_mount); freefile->fx_mode = mode; freefile->fx_oldinum = ino; freefile->fx_devvp = ip->i_devvp; LIST_INIT(&freefile->fx_jwork); UFS_LOCK(ump); ip->i_fs->fs_pendinginodes += 1; UFS_UNLOCK(ump); /* * If the inodedep does not exist, then the zero'ed inode has * been written to disk. If the allocated inode has never been * written to disk, then the on-disk inode is zero'ed. In either * case we can free the file immediately. If the journal was * canceled before being written the inode will never make it to * disk and we must send the canceled journal entrys to * ffs_freefile() to be cleared in conjunction with the bitmap. * Any blocks waiting on the inode to write can be safely freed * here as it will never been written. */ ACQUIRE_LOCK(ump); inodedep_lookup(pvp->v_mount, ino, 0, &inodedep); if (inodedep) { /* * Clear out freeblks that no longer need to reference * this inode. */ while ((freeblks = TAILQ_FIRST(&inodedep->id_freeblklst)) != NULL) { TAILQ_REMOVE(&inodedep->id_freeblklst, freeblks, fb_next); freeblks->fb_state &= ~ONDEPLIST; } /* * Remove this inode from the unlinked list. */ if (inodedep->id_state & UNLINKED) { /* * Save the journal work to be freed with the bitmap * before we clear UNLINKED. Otherwise it can be lost * if the inode block is written. */ handle_bufwait(inodedep, &freefile->fx_jwork); clear_unlinked_inodedep(inodedep); /* * Re-acquire inodedep as we've dropped the * per-filesystem lock in clear_unlinked_inodedep(). */ inodedep_lookup(pvp->v_mount, ino, 0, &inodedep); } } if (inodedep == NULL || check_inode_unwritten(inodedep)) { FREE_LOCK(ump); handle_workitem_freefile(freefile); return; } if ((inodedep->id_state & DEPCOMPLETE) == 0) inodedep->id_state |= GOINGAWAY; WORKLIST_INSERT(&inodedep->id_inowait, &freefile->fx_list); FREE_LOCK(ump); if (ip->i_number == ino) ip->i_flag |= IN_MODIFIED; } /* * Check to see if an inode has never been written to disk. If * so free the inodedep and return success, otherwise return failure. * This routine must be called with splbio interrupts blocked. * * If we still have a bitmap dependency, then the inode has never * been written to disk. Drop the dependency as it is no longer * necessary since the inode is being deallocated. We set the * ALLCOMPLETE flags since the bitmap now properly shows that the * inode is not allocated. Even if the inode is actively being * written, it has been rolled back to its zero'ed state, so we * are ensured that a zero inode is what is on the disk. For short * lived files, this change will usually result in removing all the * dependencies from the inode so that it can be freed immediately. */ static int check_inode_unwritten(inodedep) struct inodedep *inodedep; { LOCK_OWNED(VFSTOUFS(inodedep->id_list.wk_mp)); if ((inodedep->id_state & (DEPCOMPLETE | UNLINKED)) != 0 || !LIST_EMPTY(&inodedep->id_dirremhd) || !LIST_EMPTY(&inodedep->id_pendinghd) || !LIST_EMPTY(&inodedep->id_bufwait) || !LIST_EMPTY(&inodedep->id_inowait) || !TAILQ_EMPTY(&inodedep->id_inoreflst) || !TAILQ_EMPTY(&inodedep->id_inoupdt) || !TAILQ_EMPTY(&inodedep->id_newinoupdt) || !TAILQ_EMPTY(&inodedep->id_extupdt) || !TAILQ_EMPTY(&inodedep->id_newextupdt) || !TAILQ_EMPTY(&inodedep->id_freeblklst) || inodedep->id_mkdiradd != NULL || inodedep->id_nlinkdelta != 0) return (0); /* * Another process might be in initiate_write_inodeblock_ufs[12] * trying to allocate memory without holding "Softdep Lock". */ if ((inodedep->id_state & IOSTARTED) != 0 && inodedep->id_savedino1 == NULL) return (0); if (inodedep->id_state & ONDEPLIST) LIST_REMOVE(inodedep, id_deps); inodedep->id_state &= ~ONDEPLIST; inodedep->id_state |= ALLCOMPLETE; inodedep->id_bmsafemap = NULL; if (inodedep->id_state & ONWORKLIST) WORKLIST_REMOVE(&inodedep->id_list); if (inodedep->id_savedino1 != NULL) { free(inodedep->id_savedino1, M_SAVEDINO); inodedep->id_savedino1 = NULL; } if (free_inodedep(inodedep) == 0) panic("check_inode_unwritten: busy inode"); return (1); } static int check_inodedep_free(inodedep) struct inodedep *inodedep; { LOCK_OWNED(VFSTOUFS(inodedep->id_list.wk_mp)); if ((inodedep->id_state & ALLCOMPLETE) != ALLCOMPLETE || !LIST_EMPTY(&inodedep->id_dirremhd) || !LIST_EMPTY(&inodedep->id_pendinghd) || !LIST_EMPTY(&inodedep->id_bufwait) || !LIST_EMPTY(&inodedep->id_inowait) || !TAILQ_EMPTY(&inodedep->id_inoreflst) || !TAILQ_EMPTY(&inodedep->id_inoupdt) || !TAILQ_EMPTY(&inodedep->id_newinoupdt) || !TAILQ_EMPTY(&inodedep->id_extupdt) || !TAILQ_EMPTY(&inodedep->id_newextupdt) || !TAILQ_EMPTY(&inodedep->id_freeblklst) || inodedep->id_mkdiradd != NULL || inodedep->id_nlinkdelta != 0 || inodedep->id_savedino1 != NULL) return (0); return (1); } /* * Try to free an inodedep structure. Return 1 if it could be freed. */ static int free_inodedep(inodedep) struct inodedep *inodedep; { LOCK_OWNED(VFSTOUFS(inodedep->id_list.wk_mp)); if ((inodedep->id_state & (ONWORKLIST | UNLINKED)) != 0 || !check_inodedep_free(inodedep)) return (0); if (inodedep->id_state & ONDEPLIST) LIST_REMOVE(inodedep, id_deps); LIST_REMOVE(inodedep, id_hash); WORKITEM_FREE(inodedep, D_INODEDEP); return (1); } /* * Free the block referenced by a freework structure. The parent freeblks * structure is released and completed when the final cg bitmap reaches * the disk. This routine may be freeing a jnewblk which never made it to * disk in which case we do not have to wait as the operation is undone * in memory immediately. */ static void freework_freeblock(freework) struct freework *freework; { struct freeblks *freeblks; struct jnewblk *jnewblk; struct ufsmount *ump; struct workhead wkhd; struct fs *fs; int bsize; int needj; ump = VFSTOUFS(freework->fw_list.wk_mp); LOCK_OWNED(ump); /* * Handle partial truncate separately. */ if (freework->fw_indir) { complete_trunc_indir(freework); return; } freeblks = freework->fw_freeblks; fs = ump->um_fs; needj = MOUNTEDSUJ(freeblks->fb_list.wk_mp) != 0; bsize = lfragtosize(fs, freework->fw_frags); LIST_INIT(&wkhd); /* * DEPCOMPLETE is cleared in indirblk_insert() if the block lives * on the indirblk hashtable and prevents premature freeing. */ freework->fw_state |= DEPCOMPLETE; /* * SUJ needs to wait for the segment referencing freed indirect * blocks to expire so that we know the checker will not confuse * a re-allocated indirect block with its old contents. */ if (needj && freework->fw_lbn <= -NDADDR) indirblk_insert(freework); /* * If we are canceling an existing jnewblk pass it to the free * routine, otherwise pass the freeblk which will ultimately * release the freeblks. If we're not journaling, we can just * free the freeblks immediately. */ jnewblk = freework->fw_jnewblk; if (jnewblk != NULL) { cancel_jnewblk(jnewblk, &wkhd); needj = 0; } else if (needj) { freework->fw_state |= DELAYEDFREE; freeblks->fb_cgwait++; WORKLIST_INSERT(&wkhd, &freework->fw_list); } FREE_LOCK(ump); freeblks_free(ump, freeblks, btodb(bsize)); CTR4(KTR_SUJ, "freework_freeblock: ino %d blkno %jd lbn %jd size %ld", freeblks->fb_inum, freework->fw_blkno, freework->fw_lbn, bsize); ffs_blkfree(ump, fs, freeblks->fb_devvp, freework->fw_blkno, bsize, freeblks->fb_inum, freeblks->fb_vtype, &wkhd); ACQUIRE_LOCK(ump); /* * The jnewblk will be discarded and the bits in the map never * made it to disk. We can immediately free the freeblk. */ if (needj == 0) handle_written_freework(freework); } /* * We enqueue freework items that need processing back on the freeblks and * add the freeblks to the worklist. This makes it easier to find all work * required to flush a truncation in process_truncates(). */ static void freework_enqueue(freework) struct freework *freework; { struct freeblks *freeblks; freeblks = freework->fw_freeblks; if ((freework->fw_state & INPROGRESS) == 0) WORKLIST_INSERT(&freeblks->fb_freeworkhd, &freework->fw_list); if ((freeblks->fb_state & (ONWORKLIST | INPROGRESS | ALLCOMPLETE)) == ALLCOMPLETE && LIST_EMPTY(&freeblks->fb_jblkdephd)) add_to_worklist(&freeblks->fb_list, WK_NODELAY); } /* * Start, continue, or finish the process of freeing an indirect block tree. * The free operation may be paused at any point with fw_off containing the * offset to restart from. This enables us to implement some flow control * for large truncates which may fan out and generate a huge number of * dependencies. */ static void handle_workitem_indirblk(freework) struct freework *freework; { struct freeblks *freeblks; struct ufsmount *ump; struct fs *fs; freeblks = freework->fw_freeblks; ump = VFSTOUFS(freeblks->fb_list.wk_mp); fs = ump->um_fs; if (freework->fw_state & DEPCOMPLETE) { handle_written_freework(freework); return; } if (freework->fw_off == NINDIR(fs)) { freework_freeblock(freework); return; } freework->fw_state |= INPROGRESS; FREE_LOCK(ump); indir_trunc(freework, fsbtodb(fs, freework->fw_blkno), freework->fw_lbn); ACQUIRE_LOCK(ump); } /* * Called when a freework structure attached to a cg buf is written. The * ref on either the parent or the freeblks structure is released and * the freeblks is added back to the worklist if there is more work to do. */ static void handle_written_freework(freework) struct freework *freework; { struct freeblks *freeblks; struct freework *parent; freeblks = freework->fw_freeblks; parent = freework->fw_parent; if (freework->fw_state & DELAYEDFREE) freeblks->fb_cgwait--; freework->fw_state |= COMPLETE; if ((freework->fw_state & ALLCOMPLETE) == ALLCOMPLETE) WORKITEM_FREE(freework, D_FREEWORK); if (parent) { if (--parent->fw_ref == 0) freework_enqueue(parent); return; } if (--freeblks->fb_ref != 0) return; if ((freeblks->fb_state & (ALLCOMPLETE | ONWORKLIST | INPROGRESS)) == ALLCOMPLETE && LIST_EMPTY(&freeblks->fb_jblkdephd)) add_to_worklist(&freeblks->fb_list, WK_NODELAY); } /* * This workitem routine performs the block de-allocation. * The workitem is added to the pending list after the updated * inode block has been written to disk. As mentioned above, * checks regarding the number of blocks de-allocated (compared * to the number of blocks allocated for the file) are also * performed in this function. */ static int handle_workitem_freeblocks(freeblks, flags) struct freeblks *freeblks; int flags; { struct freework *freework; struct newblk *newblk; struct allocindir *aip; struct ufsmount *ump; struct worklist *wk; KASSERT(LIST_EMPTY(&freeblks->fb_jblkdephd), ("handle_workitem_freeblocks: Journal entries not written.")); ump = VFSTOUFS(freeblks->fb_list.wk_mp); ACQUIRE_LOCK(ump); while ((wk = LIST_FIRST(&freeblks->fb_freeworkhd)) != NULL) { WORKLIST_REMOVE(wk); switch (wk->wk_type) { case D_DIRREM: wk->wk_state |= COMPLETE; add_to_worklist(wk, 0); continue; case D_ALLOCDIRECT: free_newblk(WK_NEWBLK(wk)); continue; case D_ALLOCINDIR: aip = WK_ALLOCINDIR(wk); freework = NULL; if (aip->ai_state & DELAYEDFREE) { FREE_LOCK(ump); freework = newfreework(ump, freeblks, NULL, aip->ai_lbn, aip->ai_newblkno, ump->um_fs->fs_frag, 0, 0); ACQUIRE_LOCK(ump); } newblk = WK_NEWBLK(wk); if (newblk->nb_jnewblk) { freework->fw_jnewblk = newblk->nb_jnewblk; newblk->nb_jnewblk->jn_dep = &freework->fw_list; newblk->nb_jnewblk = NULL; } free_newblk(newblk); continue; case D_FREEWORK: freework = WK_FREEWORK(wk); if (freework->fw_lbn <= -NDADDR) handle_workitem_indirblk(freework); else freework_freeblock(freework); continue; default: panic("handle_workitem_freeblocks: Unknown type %s", TYPENAME(wk->wk_type)); } } if (freeblks->fb_ref != 0) { freeblks->fb_state &= ~INPROGRESS; wake_worklist(&freeblks->fb_list); freeblks = NULL; } FREE_LOCK(ump); if (freeblks) return handle_complete_freeblocks(freeblks, flags); return (0); } /* * Handle completion of block free via truncate. This allows fs_pending * to track the actual free block count more closely than if we only updated * it at the end. We must be careful to handle cases where the block count * on free was incorrect. */ static void freeblks_free(ump, freeblks, blocks) struct ufsmount *ump; struct freeblks *freeblks; int blocks; { struct fs *fs; ufs2_daddr_t remain; UFS_LOCK(ump); remain = -freeblks->fb_chkcnt; freeblks->fb_chkcnt += blocks; if (remain > 0) { if (remain < blocks) blocks = remain; fs = ump->um_fs; fs->fs_pendingblocks -= blocks; } UFS_UNLOCK(ump); } /* * Once all of the freework workitems are complete we can retire the * freeblocks dependency and any journal work awaiting completion. This * can not be called until all other dependencies are stable on disk. */ static int handle_complete_freeblocks(freeblks, flags) struct freeblks *freeblks; int flags; { struct inodedep *inodedep; struct inode *ip; struct vnode *vp; struct fs *fs; struct ufsmount *ump; ufs2_daddr_t spare; ump = VFSTOUFS(freeblks->fb_list.wk_mp); fs = ump->um_fs; flags = LK_EXCLUSIVE | flags; spare = freeblks->fb_chkcnt; /* * If we did not release the expected number of blocks we may have * to adjust the inode block count here. Only do so if it wasn't * a truncation to zero and the modrev still matches. */ if (spare && freeblks->fb_len != 0) { if (ffs_vgetf(freeblks->fb_list.wk_mp, freeblks->fb_inum, flags, &vp, FFSV_FORCEINSMQ) != 0) return (EBUSY); ip = VTOI(vp); if (DIP(ip, i_modrev) == freeblks->fb_modrev) { DIP_SET(ip, i_blocks, DIP(ip, i_blocks) - spare); ip->i_flag |= IN_CHANGE; /* * We must wait so this happens before the * journal is reclaimed. */ ffs_update(vp, 1); } vput(vp); } if (spare < 0) { UFS_LOCK(ump); fs->fs_pendingblocks += spare; UFS_UNLOCK(ump); } #ifdef QUOTA /* Handle spare. */ if (spare) quotaadj(freeblks->fb_quota, ump, -spare); quotarele(freeblks->fb_quota); #endif ACQUIRE_LOCK(ump); if (freeblks->fb_state & ONDEPLIST) { inodedep_lookup(freeblks->fb_list.wk_mp, freeblks->fb_inum, 0, &inodedep); TAILQ_REMOVE(&inodedep->id_freeblklst, freeblks, fb_next); freeblks->fb_state &= ~ONDEPLIST; if (TAILQ_EMPTY(&inodedep->id_freeblklst)) free_inodedep(inodedep); } /* * All of the freeblock deps must be complete prior to this call * so it's now safe to complete earlier outstanding journal entries. */ handle_jwork(&freeblks->fb_jwork); WORKITEM_FREE(freeblks, D_FREEBLKS); FREE_LOCK(ump); return (0); } /* * Release blocks associated with the freeblks and stored in the indirect * block dbn. If level is greater than SINGLE, the block is an indirect block * and recursive calls to indirtrunc must be used to cleanse other indirect * blocks. * * This handles partial and complete truncation of blocks. Partial is noted * with goingaway == 0. In this case the freework is completed after the * zero'd indirects are written to disk. For full truncation the freework * is completed after the block is freed. */ static void indir_trunc(freework, dbn, lbn) struct freework *freework; ufs2_daddr_t dbn; ufs_lbn_t lbn; { struct freework *nfreework; struct workhead wkhd; struct freeblks *freeblks; struct buf *bp; struct fs *fs; struct indirdep *indirdep; struct ufsmount *ump; ufs1_daddr_t *bap1 = 0; ufs2_daddr_t nb, nnb, *bap2 = 0; ufs_lbn_t lbnadd, nlbn; int i, nblocks, ufs1fmt; int freedblocks; int goingaway; int freedeps; int needj; int level; int cnt; freeblks = freework->fw_freeblks; ump = VFSTOUFS(freeblks->fb_list.wk_mp); fs = ump->um_fs; /* * Get buffer of block pointers to be freed. There are three cases: * * 1) Partial truncate caches the indirdep pointer in the freework * which provides us a back copy to the save bp which holds the * pointers we want to clear. When this completes the zero * pointers are written to the real copy. * 2) The indirect is being completely truncated, cancel_indirdep() * eliminated the real copy and placed the indirdep on the saved * copy. The indirdep and buf are discarded when this completes. * 3) The indirect was not in memory, we read a copy off of the disk * using the devvp and drop and invalidate the buffer when we're * done. */ goingaway = 1; indirdep = NULL; if (freework->fw_indir != NULL) { goingaway = 0; indirdep = freework->fw_indir; bp = indirdep->ir_savebp; if (bp == NULL || bp->b_blkno != dbn) panic("indir_trunc: Bad saved buf %p blkno %jd", bp, (intmax_t)dbn); } else if ((bp = incore(&freeblks->fb_devvp->v_bufobj, dbn)) != NULL) { /* * The lock prevents the buf dep list from changing and * indirects on devvp should only ever have one dependency. */ indirdep = WK_INDIRDEP(LIST_FIRST(&bp->b_dep)); if (indirdep == NULL || (indirdep->ir_state & GOINGAWAY) == 0) panic("indir_trunc: Bad indirdep %p from buf %p", indirdep, bp); } else if (bread(freeblks->fb_devvp, dbn, (int)fs->fs_bsize, NOCRED, &bp) != 0) { brelse(bp); return; } ACQUIRE_LOCK(ump); /* Protects against a race with complete_trunc_indir(). */ freework->fw_state &= ~INPROGRESS; /* * If we have an indirdep we need to enforce the truncation order * and discard it when it is complete. */ if (indirdep) { if (freework != TAILQ_FIRST(&indirdep->ir_trunc) && !TAILQ_EMPTY(&indirdep->ir_trunc)) { /* * Add the complete truncate to the list on the * indirdep to enforce in-order processing. */ if (freework->fw_indir == NULL) TAILQ_INSERT_TAIL(&indirdep->ir_trunc, freework, fw_next); FREE_LOCK(ump); return; } /* * If we're goingaway, free the indirdep. Otherwise it will * linger until the write completes. */ if (goingaway) free_indirdep(indirdep); } FREE_LOCK(ump); /* Initialize pointers depending on block size. */ if (ump->um_fstype == UFS1) { bap1 = (ufs1_daddr_t *)bp->b_data; nb = bap1[freework->fw_off]; ufs1fmt = 1; } else { bap2 = (ufs2_daddr_t *)bp->b_data; nb = bap2[freework->fw_off]; ufs1fmt = 0; } level = lbn_level(lbn); needj = MOUNTEDSUJ(UFSTOVFS(ump)) != 0; lbnadd = lbn_offset(fs, level); nblocks = btodb(fs->fs_bsize); nfreework = freework; freedeps = 0; cnt = 0; /* * Reclaim blocks. Traverses into nested indirect levels and * arranges for the current level to be freed when subordinates * are free when journaling. */ for (i = freework->fw_off; i < NINDIR(fs); i++, nb = nnb) { if (i != NINDIR(fs) - 1) { if (ufs1fmt) nnb = bap1[i+1]; else nnb = bap2[i+1]; } else nnb = 0; if (nb == 0) continue; cnt++; if (level != 0) { nlbn = (lbn + 1) - (i * lbnadd); if (needj != 0) { nfreework = newfreework(ump, freeblks, freework, nlbn, nb, fs->fs_frag, 0, 0); freedeps++; } indir_trunc(nfreework, fsbtodb(fs, nb), nlbn); } else { struct freedep *freedep; /* * Attempt to aggregate freedep dependencies for * all blocks being released to the same CG. */ LIST_INIT(&wkhd); if (needj != 0 && (nnb == 0 || (dtog(fs, nb) != dtog(fs, nnb)))) { freedep = newfreedep(freework); WORKLIST_INSERT_UNLOCKED(&wkhd, &freedep->fd_list); freedeps++; } CTR3(KTR_SUJ, "indir_trunc: ino %d blkno %jd size %ld", freeblks->fb_inum, nb, fs->fs_bsize); ffs_blkfree(ump, fs, freeblks->fb_devvp, nb, fs->fs_bsize, freeblks->fb_inum, freeblks->fb_vtype, &wkhd); } } if (goingaway) { bp->b_flags |= B_INVAL | B_NOCACHE; brelse(bp); } freedblocks = 0; if (level == 0) freedblocks = (nblocks * cnt); if (needj == 0) freedblocks += nblocks; freeblks_free(ump, freeblks, freedblocks); /* * If we are journaling set up the ref counts and offset so this * indirect can be completed when its children are free. */ if (needj) { ACQUIRE_LOCK(ump); freework->fw_off = i; freework->fw_ref += freedeps; freework->fw_ref -= NINDIR(fs) + 1; if (level == 0) freeblks->fb_cgwait += freedeps; if (freework->fw_ref == 0) freework_freeblock(freework); FREE_LOCK(ump); return; } /* * If we're not journaling we can free the indirect now. */ dbn = dbtofsb(fs, dbn); CTR3(KTR_SUJ, "indir_trunc 2: ino %d blkno %jd size %ld", freeblks->fb_inum, dbn, fs->fs_bsize); ffs_blkfree(ump, fs, freeblks->fb_devvp, dbn, fs->fs_bsize, freeblks->fb_inum, freeblks->fb_vtype, NULL); /* Non SUJ softdep does single-threaded truncations. */ if (freework->fw_blkno == dbn) { freework->fw_state |= ALLCOMPLETE; ACQUIRE_LOCK(ump); handle_written_freework(freework); FREE_LOCK(ump); } return; } /* * Cancel an allocindir when it is removed via truncation. When bp is not * NULL the indirect never appeared on disk and is scheduled to be freed * independently of the indir so we can more easily track journal work. */ static void cancel_allocindir(aip, bp, freeblks, trunc) struct allocindir *aip; struct buf *bp; struct freeblks *freeblks; int trunc; { struct indirdep *indirdep; struct freefrag *freefrag; struct newblk *newblk; newblk = (struct newblk *)aip; LIST_REMOVE(aip, ai_next); /* * We must eliminate the pointer in bp if it must be freed on its * own due to partial truncate or pending journal work. */ if (bp && (trunc || newblk->nb_jnewblk)) { /* * Clear the pointer and mark the aip to be freed * directly if it never existed on disk. */ aip->ai_state |= DELAYEDFREE; indirdep = aip->ai_indirdep; if (indirdep->ir_state & UFS1FMT) ((ufs1_daddr_t *)bp->b_data)[aip->ai_offset] = 0; else ((ufs2_daddr_t *)bp->b_data)[aip->ai_offset] = 0; } /* * When truncating the previous pointer will be freed via * savedbp. Eliminate the freefrag which would dup free. */ if (trunc && (freefrag = newblk->nb_freefrag) != NULL) { newblk->nb_freefrag = NULL; if (freefrag->ff_jdep) cancel_jfreefrag( WK_JFREEFRAG(freefrag->ff_jdep)); jwork_move(&freeblks->fb_jwork, &freefrag->ff_jwork); WORKITEM_FREE(freefrag, D_FREEFRAG); } /* * If the journal hasn't been written the jnewblk must be passed * to the call to ffs_blkfree that reclaims the space. We accomplish * this by leaving the journal dependency on the newblk to be freed * when a freework is created in handle_workitem_freeblocks(). */ cancel_newblk(newblk, NULL, &freeblks->fb_jwork); WORKLIST_INSERT(&freeblks->fb_freeworkhd, &newblk->nb_list); } /* * Create the mkdir dependencies for . and .. in a new directory. Link them * in to a newdirblk so any subsequent additions are tracked properly. The * caller is responsible for adding the mkdir1 dependency to the journal * and updating id_mkdiradd. This function returns with the per-filesystem * lock held. */ static struct mkdir * setup_newdir(dap, newinum, dinum, newdirbp, mkdirp) struct diradd *dap; ino_t newinum; ino_t dinum; struct buf *newdirbp; struct mkdir **mkdirp; { struct newblk *newblk; struct pagedep *pagedep; struct inodedep *inodedep; struct newdirblk *newdirblk = 0; struct mkdir *mkdir1, *mkdir2; struct worklist *wk; struct jaddref *jaddref; struct ufsmount *ump; struct mount *mp; mp = dap->da_list.wk_mp; ump = VFSTOUFS(mp); newdirblk = malloc(sizeof(struct newdirblk), M_NEWDIRBLK, M_SOFTDEP_FLAGS); workitem_alloc(&newdirblk->db_list, D_NEWDIRBLK, mp); LIST_INIT(&newdirblk->db_mkdir); mkdir1 = malloc(sizeof(struct mkdir), M_MKDIR, M_SOFTDEP_FLAGS); workitem_alloc(&mkdir1->md_list, D_MKDIR, mp); mkdir1->md_state = ATTACHED | MKDIR_BODY; mkdir1->md_diradd = dap; mkdir1->md_jaddref = NULL; mkdir2 = malloc(sizeof(struct mkdir), M_MKDIR, M_SOFTDEP_FLAGS); workitem_alloc(&mkdir2->md_list, D_MKDIR, mp); mkdir2->md_state = ATTACHED | MKDIR_PARENT; mkdir2->md_diradd = dap; mkdir2->md_jaddref = NULL; if (MOUNTEDSUJ(mp) == 0) { mkdir1->md_state |= DEPCOMPLETE; mkdir2->md_state |= DEPCOMPLETE; } /* * Dependency on "." and ".." being written to disk. */ mkdir1->md_buf = newdirbp; ACQUIRE_LOCK(VFSTOUFS(mp)); LIST_INSERT_HEAD(&ump->softdep_mkdirlisthd, mkdir1, md_mkdirs); /* * We must link the pagedep, allocdirect, and newdirblk for * the initial file page so the pointer to the new directory * is not written until the directory contents are live and * any subsequent additions are not marked live until the * block is reachable via the inode. */ if (pagedep_lookup(mp, newdirbp, newinum, 0, 0, &pagedep) == 0) panic("setup_newdir: lost pagedep"); LIST_FOREACH(wk, &newdirbp->b_dep, wk_list) if (wk->wk_type == D_ALLOCDIRECT) break; if (wk == NULL) panic("setup_newdir: lost allocdirect"); if (pagedep->pd_state & NEWBLOCK) panic("setup_newdir: NEWBLOCK already set"); newblk = WK_NEWBLK(wk); pagedep->pd_state |= NEWBLOCK; pagedep->pd_newdirblk = newdirblk; newdirblk->db_pagedep = pagedep; WORKLIST_INSERT(&newblk->nb_newdirblk, &newdirblk->db_list); WORKLIST_INSERT(&newdirblk->db_mkdir, &mkdir1->md_list); /* * Look up the inodedep for the parent directory so that we * can link mkdir2 into the pending dotdot jaddref or * the inode write if there is none. If the inode is * ALLCOMPLETE and no jaddref is present all dependencies have * been satisfied and mkdir2 can be freed. */ inodedep_lookup(mp, dinum, 0, &inodedep); if (MOUNTEDSUJ(mp)) { if (inodedep == NULL) panic("setup_newdir: Lost parent."); jaddref = (struct jaddref *)TAILQ_LAST(&inodedep->id_inoreflst, inoreflst); KASSERT(jaddref != NULL && jaddref->ja_parent == newinum && (jaddref->ja_state & MKDIR_PARENT), ("setup_newdir: bad dotdot jaddref %p", jaddref)); LIST_INSERT_HEAD(&ump->softdep_mkdirlisthd, mkdir2, md_mkdirs); mkdir2->md_jaddref = jaddref; jaddref->ja_mkdir = mkdir2; } else if (inodedep == NULL || (inodedep->id_state & ALLCOMPLETE) == ALLCOMPLETE) { dap->da_state &= ~MKDIR_PARENT; WORKITEM_FREE(mkdir2, D_MKDIR); mkdir2 = NULL; } else { LIST_INSERT_HEAD(&ump->softdep_mkdirlisthd, mkdir2, md_mkdirs); WORKLIST_INSERT(&inodedep->id_bufwait, &mkdir2->md_list); } *mkdirp = mkdir2; return (mkdir1); } /* * Directory entry addition dependencies. * * When adding a new directory entry, the inode (with its incremented link * count) must be written to disk before the directory entry's pointer to it. * Also, if the inode is newly allocated, the corresponding freemap must be * updated (on disk) before the directory entry's pointer. These requirements * are met via undo/redo on the directory entry's pointer, which consists * simply of the inode number. * * As directory entries are added and deleted, the free space within a * directory block can become fragmented. The ufs filesystem will compact * a fragmented directory block to make space for a new entry. When this * occurs, the offsets of previously added entries change. Any "diradd" * dependency structures corresponding to these entries must be updated with * the new offsets. */ /* * This routine is called after the in-memory inode's link * count has been incremented, but before the directory entry's * pointer to the inode has been set. */ int softdep_setup_directory_add(bp, dp, diroffset, newinum, newdirbp, isnewblk) struct buf *bp; /* buffer containing directory block */ struct inode *dp; /* inode for directory */ off_t diroffset; /* offset of new entry in directory */ ino_t newinum; /* inode referenced by new directory entry */ struct buf *newdirbp; /* non-NULL => contents of new mkdir */ int isnewblk; /* entry is in a newly allocated block */ { int offset; /* offset of new entry within directory block */ ufs_lbn_t lbn; /* block in directory containing new entry */ struct fs *fs; struct diradd *dap; struct newblk *newblk; struct pagedep *pagedep; struct inodedep *inodedep; struct newdirblk *newdirblk = 0; struct mkdir *mkdir1, *mkdir2; struct jaddref *jaddref; struct ufsmount *ump; struct mount *mp; int isindir; ump = dp->i_ump; mp = UFSTOVFS(ump); KASSERT(MOUNTEDSOFTDEP(mp) != 0, ("softdep_setup_directory_add called on non-softdep filesystem")); /* * Whiteouts have no dependencies. */ if (newinum == WINO) { if (newdirbp != NULL) bdwrite(newdirbp); return (0); } jaddref = NULL; mkdir1 = mkdir2 = NULL; fs = dp->i_fs; lbn = lblkno(fs, diroffset); offset = blkoff(fs, diroffset); dap = malloc(sizeof(struct diradd), M_DIRADD, M_SOFTDEP_FLAGS|M_ZERO); workitem_alloc(&dap->da_list, D_DIRADD, mp); dap->da_offset = offset; dap->da_newinum = newinum; dap->da_state = ATTACHED; LIST_INIT(&dap->da_jwork); isindir = bp->b_lblkno >= NDADDR; if (isnewblk && (isindir ? blkoff(fs, diroffset) : fragoff(fs, diroffset)) == 0) { newdirblk = malloc(sizeof(struct newdirblk), M_NEWDIRBLK, M_SOFTDEP_FLAGS); workitem_alloc(&newdirblk->db_list, D_NEWDIRBLK, mp); LIST_INIT(&newdirblk->db_mkdir); } /* * If we're creating a new directory setup the dependencies and set * the dap state to wait for them. Otherwise it's COMPLETE and * we can move on. */ if (newdirbp == NULL) { dap->da_state |= DEPCOMPLETE; ACQUIRE_LOCK(ump); } else { dap->da_state |= MKDIR_BODY | MKDIR_PARENT; mkdir1 = setup_newdir(dap, newinum, dp->i_number, newdirbp, &mkdir2); } /* * Link into parent directory pagedep to await its being written. */ pagedep_lookup(mp, bp, dp->i_number, lbn, DEPALLOC, &pagedep); #ifdef DEBUG if (diradd_lookup(pagedep, offset) != NULL) panic("softdep_setup_directory_add: %p already at off %d\n", diradd_lookup(pagedep, offset), offset); #endif dap->da_pagedep = pagedep; LIST_INSERT_HEAD(&pagedep->pd_diraddhd[DIRADDHASH(offset)], dap, da_pdlist); inodedep_lookup(mp, newinum, DEPALLOC, &inodedep); /* * If we're journaling, link the diradd into the jaddref so it * may be completed after the journal entry is written. Otherwise, * link the diradd into its inodedep. If the inode is not yet * written place it on the bufwait list, otherwise do the post-inode * write processing to put it on the id_pendinghd list. */ if (MOUNTEDSUJ(mp)) { jaddref = (struct jaddref *)TAILQ_LAST(&inodedep->id_inoreflst, inoreflst); KASSERT(jaddref != NULL && jaddref->ja_parent == dp->i_number, ("softdep_setup_directory_add: bad jaddref %p", jaddref)); jaddref->ja_diroff = diroffset; jaddref->ja_diradd = dap; add_to_journal(&jaddref->ja_list); } else if ((inodedep->id_state & ALLCOMPLETE) == ALLCOMPLETE) diradd_inode_written(dap, inodedep); else WORKLIST_INSERT(&inodedep->id_bufwait, &dap->da_list); /* * Add the journal entries for . and .. links now that the primary * link is written. */ if (mkdir1 != NULL && MOUNTEDSUJ(mp)) { jaddref = (struct jaddref *)TAILQ_PREV(&jaddref->ja_ref, inoreflst, if_deps); KASSERT(jaddref != NULL && jaddref->ja_ino == jaddref->ja_parent && (jaddref->ja_state & MKDIR_BODY), ("softdep_setup_directory_add: bad dot jaddref %p", jaddref)); mkdir1->md_jaddref = jaddref; jaddref->ja_mkdir = mkdir1; /* * It is important that the dotdot journal entry * is added prior to the dot entry since dot writes * both the dot and dotdot links. These both must * be added after the primary link for the journal * to remain consistent. */ add_to_journal(&mkdir2->md_jaddref->ja_list); add_to_journal(&jaddref->ja_list); } /* * If we are adding a new directory remember this diradd so that if * we rename it we can keep the dot and dotdot dependencies. If * we are adding a new name for an inode that has a mkdiradd we * must be in rename and we have to move the dot and dotdot * dependencies to this new name. The old name is being orphaned * soon. */ if (mkdir1 != NULL) { if (inodedep->id_mkdiradd != NULL) panic("softdep_setup_directory_add: Existing mkdir"); inodedep->id_mkdiradd = dap; } else if (inodedep->id_mkdiradd) merge_diradd(inodedep, dap); if (newdirblk) { /* * There is nothing to do if we are already tracking * this block. */ if ((pagedep->pd_state & NEWBLOCK) != 0) { WORKITEM_FREE(newdirblk, D_NEWDIRBLK); FREE_LOCK(ump); return (0); } if (newblk_lookup(mp, dbtofsb(fs, bp->b_blkno), 0, &newblk) == 0) panic("softdep_setup_directory_add: lost entry"); WORKLIST_INSERT(&newblk->nb_newdirblk, &newdirblk->db_list); pagedep->pd_state |= NEWBLOCK; pagedep->pd_newdirblk = newdirblk; newdirblk->db_pagedep = pagedep; FREE_LOCK(ump); /* * If we extended into an indirect signal direnter to sync. */ if (isindir) return (1); return (0); } FREE_LOCK(ump); return (0); } /* * This procedure is called to change the offset of a directory * entry when compacting a directory block which must be owned * exclusively by the caller. Note that the actual entry movement * must be done in this procedure to ensure that no I/O completions * occur while the move is in progress. */ void softdep_change_directoryentry_offset(bp, dp, base, oldloc, newloc, entrysize) struct buf *bp; /* Buffer holding directory block. */ struct inode *dp; /* inode for directory */ caddr_t base; /* address of dp->i_offset */ caddr_t oldloc; /* address of old directory location */ caddr_t newloc; /* address of new directory location */ int entrysize; /* size of directory entry */ { int offset, oldoffset, newoffset; struct pagedep *pagedep; struct jmvref *jmvref; struct diradd *dap; struct direct *de; struct mount *mp; ufs_lbn_t lbn; int flags; mp = UFSTOVFS(dp->i_ump); KASSERT(MOUNTEDSOFTDEP(mp) != 0, ("softdep_change_directoryentry_offset called on " "non-softdep filesystem")); de = (struct direct *)oldloc; jmvref = NULL; flags = 0; /* * Moves are always journaled as it would be too complex to * determine if any affected adds or removes are present in the * journal. */ if (MOUNTEDSUJ(mp)) { flags = DEPALLOC; jmvref = newjmvref(dp, de->d_ino, dp->i_offset + (oldloc - base), dp->i_offset + (newloc - base)); } lbn = lblkno(dp->i_fs, dp->i_offset); offset = blkoff(dp->i_fs, dp->i_offset); oldoffset = offset + (oldloc - base); newoffset = offset + (newloc - base); ACQUIRE_LOCK(dp->i_ump); if (pagedep_lookup(mp, bp, dp->i_number, lbn, flags, &pagedep) == 0) goto done; dap = diradd_lookup(pagedep, oldoffset); if (dap) { dap->da_offset = newoffset; newoffset = DIRADDHASH(newoffset); oldoffset = DIRADDHASH(oldoffset); if ((dap->da_state & ALLCOMPLETE) != ALLCOMPLETE && newoffset != oldoffset) { LIST_REMOVE(dap, da_pdlist); LIST_INSERT_HEAD(&pagedep->pd_diraddhd[newoffset], dap, da_pdlist); } } done: if (jmvref) { jmvref->jm_pagedep = pagedep; LIST_INSERT_HEAD(&pagedep->pd_jmvrefhd, jmvref, jm_deps); add_to_journal(&jmvref->jm_list); } bcopy(oldloc, newloc, entrysize); FREE_LOCK(dp->i_ump); } /* * Move the mkdir dependencies and journal work from one diradd to another * when renaming a directory. The new name must depend on the mkdir deps * completing as the old name did. Directories can only have one valid link * at a time so one must be canonical. */ static void merge_diradd(inodedep, newdap) struct inodedep *inodedep; struct diradd *newdap; { struct diradd *olddap; struct mkdir *mkdir, *nextmd; struct ufsmount *ump; short state; olddap = inodedep->id_mkdiradd; inodedep->id_mkdiradd = newdap; if ((olddap->da_state & (MKDIR_PARENT | MKDIR_BODY)) != 0) { newdap->da_state &= ~DEPCOMPLETE; ump = VFSTOUFS(inodedep->id_list.wk_mp); for (mkdir = LIST_FIRST(&ump->softdep_mkdirlisthd); mkdir; mkdir = nextmd) { nextmd = LIST_NEXT(mkdir, md_mkdirs); if (mkdir->md_diradd != olddap) continue; mkdir->md_diradd = newdap; state = mkdir->md_state & (MKDIR_PARENT | MKDIR_BODY); newdap->da_state |= state; olddap->da_state &= ~state; if ((olddap->da_state & (MKDIR_PARENT | MKDIR_BODY)) == 0) break; } if ((olddap->da_state & (MKDIR_PARENT | MKDIR_BODY)) != 0) panic("merge_diradd: unfound ref"); } /* * Any mkdir related journal items are not safe to be freed until * the new name is stable. */ jwork_move(&newdap->da_jwork, &olddap->da_jwork); olddap->da_state |= DEPCOMPLETE; complete_diradd(olddap); } /* * Move the diradd to the pending list when all diradd dependencies are * complete. */ static void complete_diradd(dap) struct diradd *dap; { struct pagedep *pagedep; if ((dap->da_state & ALLCOMPLETE) == ALLCOMPLETE) { if (dap->da_state & DIRCHG) pagedep = dap->da_previous->dm_pagedep; else pagedep = dap->da_pagedep; LIST_REMOVE(dap, da_pdlist); LIST_INSERT_HEAD(&pagedep->pd_pendinghd, dap, da_pdlist); } } /* * Cancel a diradd when a dirrem overlaps with it. We must cancel the journal * add entries and conditonally journal the remove. */ static void cancel_diradd(dap, dirrem, jremref, dotremref, dotdotremref) struct diradd *dap; struct dirrem *dirrem; struct jremref *jremref; struct jremref *dotremref; struct jremref *dotdotremref; { struct inodedep *inodedep; struct jaddref *jaddref; struct inoref *inoref; struct ufsmount *ump; struct mkdir *mkdir; /* * If no remove references were allocated we're on a non-journaled * filesystem and can skip the cancel step. */ if (jremref == NULL) { free_diradd(dap, NULL); return; } /* * Cancel the primary name an free it if it does not require * journaling. */ if (inodedep_lookup(dap->da_list.wk_mp, dap->da_newinum, 0, &inodedep) != 0) { /* Abort the addref that reference this diradd. */ TAILQ_FOREACH(inoref, &inodedep->id_inoreflst, if_deps) { if (inoref->if_list.wk_type != D_JADDREF) continue; jaddref = (struct jaddref *)inoref; if (jaddref->ja_diradd != dap) continue; if (cancel_jaddref(jaddref, inodedep, &dirrem->dm_jwork) == 0) { free_jremref(jremref); jremref = NULL; } break; } } /* * Cancel subordinate names and free them if they do not require * journaling. */ if ((dap->da_state & (MKDIR_PARENT | MKDIR_BODY)) != 0) { ump = VFSTOUFS(dap->da_list.wk_mp); LIST_FOREACH(mkdir, &ump->softdep_mkdirlisthd, md_mkdirs) { if (mkdir->md_diradd != dap) continue; if ((jaddref = mkdir->md_jaddref) == NULL) continue; mkdir->md_jaddref = NULL; if (mkdir->md_state & MKDIR_PARENT) { if (cancel_jaddref(jaddref, NULL, &dirrem->dm_jwork) == 0) { free_jremref(dotdotremref); dotdotremref = NULL; } } else { if (cancel_jaddref(jaddref, inodedep, &dirrem->dm_jwork) == 0) { free_jremref(dotremref); dotremref = NULL; } } } } if (jremref) journal_jremref(dirrem, jremref, inodedep); if (dotremref) journal_jremref(dirrem, dotremref, inodedep); if (dotdotremref) journal_jremref(dirrem, dotdotremref, NULL); jwork_move(&dirrem->dm_jwork, &dap->da_jwork); free_diradd(dap, &dirrem->dm_jwork); } /* * Free a diradd dependency structure. This routine must be called * with splbio interrupts blocked. */ static void free_diradd(dap, wkhd) struct diradd *dap; struct workhead *wkhd; { struct dirrem *dirrem; struct pagedep *pagedep; struct inodedep *inodedep; struct mkdir *mkdir, *nextmd; struct ufsmount *ump; ump = VFSTOUFS(dap->da_list.wk_mp); LOCK_OWNED(ump); LIST_REMOVE(dap, da_pdlist); if (dap->da_state & ONWORKLIST) WORKLIST_REMOVE(&dap->da_list); if ((dap->da_state & DIRCHG) == 0) { pagedep = dap->da_pagedep; } else { dirrem = dap->da_previous; pagedep = dirrem->dm_pagedep; dirrem->dm_dirinum = pagedep->pd_ino; dirrem->dm_state |= COMPLETE; if (LIST_EMPTY(&dirrem->dm_jremrefhd)) add_to_worklist(&dirrem->dm_list, 0); } if (inodedep_lookup(pagedep->pd_list.wk_mp, dap->da_newinum, 0, &inodedep) != 0) if (inodedep->id_mkdiradd == dap) inodedep->id_mkdiradd = NULL; if ((dap->da_state & (MKDIR_PARENT | MKDIR_BODY)) != 0) { for (mkdir = LIST_FIRST(&ump->softdep_mkdirlisthd); mkdir; mkdir = nextmd) { nextmd = LIST_NEXT(mkdir, md_mkdirs); if (mkdir->md_diradd != dap) continue; dap->da_state &= ~(mkdir->md_state & (MKDIR_PARENT | MKDIR_BODY)); LIST_REMOVE(mkdir, md_mkdirs); if (mkdir->md_state & ONWORKLIST) WORKLIST_REMOVE(&mkdir->md_list); if (mkdir->md_jaddref != NULL) panic("free_diradd: Unexpected jaddref"); WORKITEM_FREE(mkdir, D_MKDIR); if ((dap->da_state & (MKDIR_PARENT | MKDIR_BODY)) == 0) break; } if ((dap->da_state & (MKDIR_PARENT | MKDIR_BODY)) != 0) panic("free_diradd: unfound ref"); } if (inodedep) free_inodedep(inodedep); /* * Free any journal segments waiting for the directory write. */ handle_jwork(&dap->da_jwork); WORKITEM_FREE(dap, D_DIRADD); } /* * Directory entry removal dependencies. * * When removing a directory entry, the entry's inode pointer must be * zero'ed on disk before the corresponding inode's link count is decremented * (possibly freeing the inode for re-use). This dependency is handled by * updating the directory entry but delaying the inode count reduction until * after the directory block has been written to disk. After this point, the * inode count can be decremented whenever it is convenient. */ /* * This routine should be called immediately after removing * a directory entry. The inode's link count should not be * decremented by the calling procedure -- the soft updates * code will do this task when it is safe. */ void softdep_setup_remove(bp, dp, ip, isrmdir) struct buf *bp; /* buffer containing directory block */ struct inode *dp; /* inode for the directory being modified */ struct inode *ip; /* inode for directory entry being removed */ int isrmdir; /* indicates if doing RMDIR */ { struct dirrem *dirrem, *prevdirrem; struct inodedep *inodedep; int direct; KASSERT(MOUNTEDSOFTDEP(UFSTOVFS(ip->i_ump)) != 0, ("softdep_setup_remove called on non-softdep filesystem")); /* * Allocate a new dirrem if appropriate and ACQUIRE_LOCK. We want * newdirrem() to setup the full directory remove which requires * isrmdir > 1. */ dirrem = newdirrem(bp, dp, ip, isrmdir, &prevdirrem); /* * Add the dirrem to the inodedep's pending remove list for quick * discovery later. */ if (inodedep_lookup(UFSTOVFS(ip->i_ump), ip->i_number, 0, &inodedep) == 0) panic("softdep_setup_remove: Lost inodedep."); KASSERT((inodedep->id_state & UNLINKED) == 0, ("inode unlinked")); dirrem->dm_state |= ONDEPLIST; LIST_INSERT_HEAD(&inodedep->id_dirremhd, dirrem, dm_inonext); /* * If the COMPLETE flag is clear, then there were no active * entries and we want to roll back to a zeroed entry until * the new inode is committed to disk. If the COMPLETE flag is * set then we have deleted an entry that never made it to * disk. If the entry we deleted resulted from a name change, * then the old name still resides on disk. We cannot delete * its inode (returned to us in prevdirrem) until the zeroed * directory entry gets to disk. The new inode has never been * referenced on the disk, so can be deleted immediately. */ if ((dirrem->dm_state & COMPLETE) == 0) { LIST_INSERT_HEAD(&dirrem->dm_pagedep->pd_dirremhd, dirrem, dm_next); FREE_LOCK(ip->i_ump); } else { if (prevdirrem != NULL) LIST_INSERT_HEAD(&dirrem->dm_pagedep->pd_dirremhd, prevdirrem, dm_next); dirrem->dm_dirinum = dirrem->dm_pagedep->pd_ino; direct = LIST_EMPTY(&dirrem->dm_jremrefhd); FREE_LOCK(ip->i_ump); if (direct) handle_workitem_remove(dirrem, 0); } } /* * Check for an entry matching 'offset' on both the pd_dirraddhd list and the * pd_pendinghd list of a pagedep. */ static struct diradd * diradd_lookup(pagedep, offset) struct pagedep *pagedep; int offset; { struct diradd *dap; LIST_FOREACH(dap, &pagedep->pd_diraddhd[DIRADDHASH(offset)], da_pdlist) if (dap->da_offset == offset) return (dap); LIST_FOREACH(dap, &pagedep->pd_pendinghd, da_pdlist) if (dap->da_offset == offset) return (dap); return (NULL); } /* * Search for a .. diradd dependency in a directory that is being removed. * If the directory was renamed to a new parent we have a diradd rather * than a mkdir for the .. entry. We need to cancel it now before * it is found in truncate(). */ static struct jremref * cancel_diradd_dotdot(ip, dirrem, jremref) struct inode *ip; struct dirrem *dirrem; struct jremref *jremref; { struct pagedep *pagedep; struct diradd *dap; struct worklist *wk; if (pagedep_lookup(UFSTOVFS(ip->i_ump), NULL, ip->i_number, 0, 0, &pagedep) == 0) return (jremref); dap = diradd_lookup(pagedep, DOTDOT_OFFSET); if (dap == NULL) return (jremref); cancel_diradd(dap, dirrem, jremref, NULL, NULL); /* * Mark any journal work as belonging to the parent so it is freed * with the .. reference. */ LIST_FOREACH(wk, &dirrem->dm_jwork, wk_list) wk->wk_state |= MKDIR_PARENT; return (NULL); } /* * Cancel the MKDIR_PARENT mkdir component of a diradd when we're going to * replace it with a dirrem/diradd pair as a result of re-parenting a * directory. This ensures that we don't simultaneously have a mkdir and * a diradd for the same .. entry. */ static struct jremref * cancel_mkdir_dotdot(ip, dirrem, jremref) struct inode *ip; struct dirrem *dirrem; struct jremref *jremref; { struct inodedep *inodedep; struct jaddref *jaddref; struct ufsmount *ump; struct mkdir *mkdir; struct diradd *dap; if (inodedep_lookup(UFSTOVFS(ip->i_ump), ip->i_number, 0, &inodedep) == 0) return (jremref); dap = inodedep->id_mkdiradd; if (dap == NULL || (dap->da_state & MKDIR_PARENT) == 0) return (jremref); ump = VFSTOUFS(inodedep->id_list.wk_mp); for (mkdir = LIST_FIRST(&ump->softdep_mkdirlisthd); mkdir; mkdir = LIST_NEXT(mkdir, md_mkdirs)) if (mkdir->md_diradd == dap && mkdir->md_state & MKDIR_PARENT) break; if (mkdir == NULL) panic("cancel_mkdir_dotdot: Unable to find mkdir\n"); if ((jaddref = mkdir->md_jaddref) != NULL) { mkdir->md_jaddref = NULL; jaddref->ja_state &= ~MKDIR_PARENT; if (inodedep_lookup(UFSTOVFS(ip->i_ump), jaddref->ja_ino, 0, &inodedep) == 0) panic("cancel_mkdir_dotdot: Lost parent inodedep"); if (cancel_jaddref(jaddref, inodedep, &dirrem->dm_jwork)) { journal_jremref(dirrem, jremref, inodedep); jremref = NULL; } } if (mkdir->md_state & ONWORKLIST) WORKLIST_REMOVE(&mkdir->md_list); mkdir->md_state |= ALLCOMPLETE; complete_mkdir(mkdir); return (jremref); } static void journal_jremref(dirrem, jremref, inodedep) struct dirrem *dirrem; struct jremref *jremref; struct inodedep *inodedep; { if (inodedep == NULL) if (inodedep_lookup(jremref->jr_list.wk_mp, jremref->jr_ref.if_ino, 0, &inodedep) == 0) panic("journal_jremref: Lost inodedep"); LIST_INSERT_HEAD(&dirrem->dm_jremrefhd, jremref, jr_deps); TAILQ_INSERT_TAIL(&inodedep->id_inoreflst, &jremref->jr_ref, if_deps); add_to_journal(&jremref->jr_list); } static void dirrem_journal(dirrem, jremref, dotremref, dotdotremref) struct dirrem *dirrem; struct jremref *jremref; struct jremref *dotremref; struct jremref *dotdotremref; { struct inodedep *inodedep; if (inodedep_lookup(jremref->jr_list.wk_mp, jremref->jr_ref.if_ino, 0, &inodedep) == 0) panic("dirrem_journal: Lost inodedep"); journal_jremref(dirrem, jremref, inodedep); if (dotremref) journal_jremref(dirrem, dotremref, inodedep); if (dotdotremref) journal_jremref(dirrem, dotdotremref, NULL); } /* * Allocate a new dirrem if appropriate and return it along with * its associated pagedep. Called without a lock, returns with lock. */ static struct dirrem * newdirrem(bp, dp, ip, isrmdir, prevdirremp) struct buf *bp; /* buffer containing directory block */ struct inode *dp; /* inode for the directory being modified */ struct inode *ip; /* inode for directory entry being removed */ int isrmdir; /* indicates if doing RMDIR */ struct dirrem **prevdirremp; /* previously referenced inode, if any */ { int offset; ufs_lbn_t lbn; struct diradd *dap; struct dirrem *dirrem; struct pagedep *pagedep; struct jremref *jremref; struct jremref *dotremref; struct jremref *dotdotremref; struct vnode *dvp; /* * Whiteouts have no deletion dependencies. */ if (ip == NULL) panic("newdirrem: whiteout"); dvp = ITOV(dp); /* * If the system is over its limit and our filesystem is * responsible for more than our share of that usage and * we are not a snapshot, request some inodedep cleanup. * Limiting the number of dirrem structures will also limit * the number of freefile and freeblks structures. */ ACQUIRE_LOCK(ip->i_ump); if (!IS_SNAPSHOT(ip) && softdep_excess_dirrem(ip->i_ump)) schedule_cleanup(ITOV(dp)->v_mount); else FREE_LOCK(ip->i_ump); dirrem = malloc(sizeof(struct dirrem), M_DIRREM, M_SOFTDEP_FLAGS | M_ZERO); workitem_alloc(&dirrem->dm_list, D_DIRREM, dvp->v_mount); LIST_INIT(&dirrem->dm_jremrefhd); LIST_INIT(&dirrem->dm_jwork); dirrem->dm_state = isrmdir ? RMDIR : 0; dirrem->dm_oldinum = ip->i_number; *prevdirremp = NULL; /* * Allocate remove reference structures to track journal write * dependencies. We will always have one for the link and * when doing directories we will always have one more for dot. * When renaming a directory we skip the dotdot link change so * this is not needed. */ jremref = dotremref = dotdotremref = NULL; if (DOINGSUJ(dvp)) { if (isrmdir) { jremref = newjremref(dirrem, dp, ip, dp->i_offset, ip->i_effnlink + 2); dotremref = newjremref(dirrem, ip, ip, DOT_OFFSET, ip->i_effnlink + 1); dotdotremref = newjremref(dirrem, ip, dp, DOTDOT_OFFSET, dp->i_effnlink + 1); dotdotremref->jr_state |= MKDIR_PARENT; } else jremref = newjremref(dirrem, dp, ip, dp->i_offset, ip->i_effnlink + 1); } ACQUIRE_LOCK(ip->i_ump); lbn = lblkno(dp->i_fs, dp->i_offset); offset = blkoff(dp->i_fs, dp->i_offset); pagedep_lookup(UFSTOVFS(dp->i_ump), bp, dp->i_number, lbn, DEPALLOC, &pagedep); dirrem->dm_pagedep = pagedep; dirrem->dm_offset = offset; /* * If we're renaming a .. link to a new directory, cancel any * existing MKDIR_PARENT mkdir. If it has already been canceled * the jremref is preserved for any potential diradd in this * location. This can not coincide with a rmdir. */ if (dp->i_offset == DOTDOT_OFFSET) { if (isrmdir) panic("newdirrem: .. directory change during remove?"); jremref = cancel_mkdir_dotdot(dp, dirrem, jremref); } /* * If we're removing a directory search for the .. dependency now and * cancel it. Any pending journal work will be added to the dirrem * to be completed when the workitem remove completes. */ if (isrmdir) dotdotremref = cancel_diradd_dotdot(ip, dirrem, dotdotremref); /* * Check for a diradd dependency for the same directory entry. * If present, then both dependencies become obsolete and can * be de-allocated. */ dap = diradd_lookup(pagedep, offset); if (dap == NULL) { /* * Link the jremref structures into the dirrem so they are * written prior to the pagedep. */ if (jremref) dirrem_journal(dirrem, jremref, dotremref, dotdotremref); return (dirrem); } /* * Must be ATTACHED at this point. */ if ((dap->da_state & ATTACHED) == 0) panic("newdirrem: not ATTACHED"); if (dap->da_newinum != ip->i_number) panic("newdirrem: inum %ju should be %ju", (uintmax_t)ip->i_number, (uintmax_t)dap->da_newinum); /* * If we are deleting a changed name that never made it to disk, * then return the dirrem describing the previous inode (which * represents the inode currently referenced from this entry on disk). */ if ((dap->da_state & DIRCHG) != 0) { *prevdirremp = dap->da_previous; dap->da_state &= ~DIRCHG; dap->da_pagedep = pagedep; } /* * We are deleting an entry that never made it to disk. * Mark it COMPLETE so we can delete its inode immediately. */ dirrem->dm_state |= COMPLETE; cancel_diradd(dap, dirrem, jremref, dotremref, dotdotremref); #ifdef SUJ_DEBUG if (isrmdir == 0) { struct worklist *wk; LIST_FOREACH(wk, &dirrem->dm_jwork, wk_list) if (wk->wk_state & (MKDIR_BODY | MKDIR_PARENT)) panic("bad wk %p (0x%X)\n", wk, wk->wk_state); } #endif return (dirrem); } /* * Directory entry change dependencies. * * Changing an existing directory entry requires that an add operation * be completed first followed by a deletion. The semantics for the addition * are identical to the description of adding a new entry above except * that the rollback is to the old inode number rather than zero. Once * the addition dependency is completed, the removal is done as described * in the removal routine above. */ /* * This routine should be called immediately after changing * a directory entry. The inode's link count should not be * decremented by the calling procedure -- the soft updates * code will perform this task when it is safe. */ void softdep_setup_directory_change(bp, dp, ip, newinum, isrmdir) struct buf *bp; /* buffer containing directory block */ struct inode *dp; /* inode for the directory being modified */ struct inode *ip; /* inode for directory entry being removed */ ino_t newinum; /* new inode number for changed entry */ int isrmdir; /* indicates if doing RMDIR */ { int offset; struct diradd *dap = NULL; struct dirrem *dirrem, *prevdirrem; struct pagedep *pagedep; struct inodedep *inodedep; struct jaddref *jaddref; struct mount *mp; offset = blkoff(dp->i_fs, dp->i_offset); mp = UFSTOVFS(dp->i_ump); KASSERT(MOUNTEDSOFTDEP(mp) != 0, ("softdep_setup_directory_change called on non-softdep filesystem")); /* * Whiteouts do not need diradd dependencies. */ if (newinum != WINO) { dap = malloc(sizeof(struct diradd), M_DIRADD, M_SOFTDEP_FLAGS|M_ZERO); workitem_alloc(&dap->da_list, D_DIRADD, mp); dap->da_state = DIRCHG | ATTACHED | DEPCOMPLETE; dap->da_offset = offset; dap->da_newinum = newinum; LIST_INIT(&dap->da_jwork); } /* * Allocate a new dirrem and ACQUIRE_LOCK. */ dirrem = newdirrem(bp, dp, ip, isrmdir, &prevdirrem); pagedep = dirrem->dm_pagedep; /* * The possible values for isrmdir: * 0 - non-directory file rename * 1 - directory rename within same directory * inum - directory rename to new directory of given inode number * When renaming to a new directory, we are both deleting and * creating a new directory entry, so the link count on the new * directory should not change. Thus we do not need the followup * dirrem which is usually done in handle_workitem_remove. We set * the DIRCHG flag to tell handle_workitem_remove to skip the * followup dirrem. */ if (isrmdir > 1) dirrem->dm_state |= DIRCHG; /* * Whiteouts have no additional dependencies, * so just put the dirrem on the correct list. */ if (newinum == WINO) { if ((dirrem->dm_state & COMPLETE) == 0) { LIST_INSERT_HEAD(&pagedep->pd_dirremhd, dirrem, dm_next); } else { dirrem->dm_dirinum = pagedep->pd_ino; if (LIST_EMPTY(&dirrem->dm_jremrefhd)) add_to_worklist(&dirrem->dm_list, 0); } FREE_LOCK(dp->i_ump); return; } /* * Add the dirrem to the inodedep's pending remove list for quick * discovery later. A valid nlinkdelta ensures that this lookup * will not fail. */ if (inodedep_lookup(mp, ip->i_number, 0, &inodedep) == 0) panic("softdep_setup_directory_change: Lost inodedep."); dirrem->dm_state |= ONDEPLIST; LIST_INSERT_HEAD(&inodedep->id_dirremhd, dirrem, dm_inonext); /* * If the COMPLETE flag is clear, then there were no active * entries and we want to roll back to the previous inode until * the new inode is committed to disk. If the COMPLETE flag is * set, then we have deleted an entry that never made it to disk. * If the entry we deleted resulted from a name change, then the old * inode reference still resides on disk. Any rollback that we do * needs to be to that old inode (returned to us in prevdirrem). If * the entry we deleted resulted from a create, then there is * no entry on the disk, so we want to roll back to zero rather * than the uncommitted inode. In either of the COMPLETE cases we * want to immediately free the unwritten and unreferenced inode. */ if ((dirrem->dm_state & COMPLETE) == 0) { dap->da_previous = dirrem; } else { if (prevdirrem != NULL) { dap->da_previous = prevdirrem; } else { dap->da_state &= ~DIRCHG; dap->da_pagedep = pagedep; } dirrem->dm_dirinum = pagedep->pd_ino; if (LIST_EMPTY(&dirrem->dm_jremrefhd)) add_to_worklist(&dirrem->dm_list, 0); } /* * Lookup the jaddref for this journal entry. We must finish * initializing it and make the diradd write dependent on it. * If we're not journaling, put it on the id_bufwait list if the * inode is not yet written. If it is written, do the post-inode * write processing to put it on the id_pendinghd list. */ inodedep_lookup(mp, newinum, DEPALLOC, &inodedep); if (MOUNTEDSUJ(mp)) { jaddref = (struct jaddref *)TAILQ_LAST(&inodedep->id_inoreflst, inoreflst); KASSERT(jaddref != NULL && jaddref->ja_parent == dp->i_number, ("softdep_setup_directory_change: bad jaddref %p", jaddref)); jaddref->ja_diroff = dp->i_offset; jaddref->ja_diradd = dap; LIST_INSERT_HEAD(&pagedep->pd_diraddhd[DIRADDHASH(offset)], dap, da_pdlist); add_to_journal(&jaddref->ja_list); } else if ((inodedep->id_state & ALLCOMPLETE) == ALLCOMPLETE) { dap->da_state |= COMPLETE; LIST_INSERT_HEAD(&pagedep->pd_pendinghd, dap, da_pdlist); WORKLIST_INSERT(&inodedep->id_pendinghd, &dap->da_list); } else { LIST_INSERT_HEAD(&pagedep->pd_diraddhd[DIRADDHASH(offset)], dap, da_pdlist); WORKLIST_INSERT(&inodedep->id_bufwait, &dap->da_list); } /* * If we're making a new name for a directory that has not been * committed when need to move the dot and dotdot references to * this new name. */ if (inodedep->id_mkdiradd && dp->i_offset != DOTDOT_OFFSET) merge_diradd(inodedep, dap); FREE_LOCK(dp->i_ump); } /* * Called whenever the link count on an inode is changed. * It creates an inode dependency so that the new reference(s) * to the inode cannot be committed to disk until the updated * inode has been written. */ void softdep_change_linkcnt(ip) struct inode *ip; /* the inode with the increased link count */ { struct inodedep *inodedep; KASSERT(MOUNTEDSOFTDEP(UFSTOVFS(ip->i_ump)) != 0, ("softdep_change_linkcnt called on non-softdep filesystem")); ACQUIRE_LOCK(ip->i_ump); inodedep_lookup(UFSTOVFS(ip->i_ump), ip->i_number, DEPALLOC, &inodedep); if (ip->i_nlink < ip->i_effnlink) panic("softdep_change_linkcnt: bad delta"); inodedep->id_nlinkdelta = ip->i_nlink - ip->i_effnlink; FREE_LOCK(ip->i_ump); } /* * Attach a sbdep dependency to the superblock buf so that we can keep * track of the head of the linked list of referenced but unlinked inodes. */ void softdep_setup_sbupdate(ump, fs, bp) struct ufsmount *ump; struct fs *fs; struct buf *bp; { struct sbdep *sbdep; struct worklist *wk; KASSERT(MOUNTEDSOFTDEP(UFSTOVFS(ump)) != 0, ("softdep_setup_sbupdate called on non-softdep filesystem")); LIST_FOREACH(wk, &bp->b_dep, wk_list) if (wk->wk_type == D_SBDEP) break; if (wk != NULL) return; sbdep = malloc(sizeof(struct sbdep), M_SBDEP, M_SOFTDEP_FLAGS); workitem_alloc(&sbdep->sb_list, D_SBDEP, UFSTOVFS(ump)); sbdep->sb_fs = fs; sbdep->sb_ump = ump; ACQUIRE_LOCK(ump); WORKLIST_INSERT(&bp->b_dep, &sbdep->sb_list); FREE_LOCK(ump); } /* * Return the first unlinked inodedep which is ready to be the head of the * list. The inodedep and all those after it must have valid next pointers. */ static struct inodedep * first_unlinked_inodedep(ump) struct ufsmount *ump; { struct inodedep *inodedep; struct inodedep *idp; LOCK_OWNED(ump); for (inodedep = TAILQ_LAST(&ump->softdep_unlinked, inodedeplst); inodedep; inodedep = idp) { if ((inodedep->id_state & UNLINKNEXT) == 0) return (NULL); idp = TAILQ_PREV(inodedep, inodedeplst, id_unlinked); if (idp == NULL || (idp->id_state & UNLINKNEXT) == 0) break; if ((inodedep->id_state & UNLINKPREV) == 0) break; } return (inodedep); } /* * Set the sujfree unlinked head pointer prior to writing a superblock. */ static void initiate_write_sbdep(sbdep) struct sbdep *sbdep; { struct inodedep *inodedep; struct fs *bpfs; struct fs *fs; bpfs = sbdep->sb_fs; fs = sbdep->sb_ump->um_fs; inodedep = first_unlinked_inodedep(sbdep->sb_ump); if (inodedep) { fs->fs_sujfree = inodedep->id_ino; inodedep->id_state |= UNLINKPREV; } else fs->fs_sujfree = 0; bpfs->fs_sujfree = fs->fs_sujfree; } /* * After a superblock is written determine whether it must be written again * due to a changing unlinked list head. */ static int handle_written_sbdep(sbdep, bp) struct sbdep *sbdep; struct buf *bp; { struct inodedep *inodedep; struct fs *fs; LOCK_OWNED(sbdep->sb_ump); fs = sbdep->sb_fs; /* * If the superblock doesn't match the in-memory list start over. */ inodedep = first_unlinked_inodedep(sbdep->sb_ump); if ((inodedep && fs->fs_sujfree != inodedep->id_ino) || (inodedep == NULL && fs->fs_sujfree != 0)) { bdirty(bp); return (1); } WORKITEM_FREE(sbdep, D_SBDEP); if (fs->fs_sujfree == 0) return (0); /* * Now that we have a record of this inode in stable store allow it * to be written to free up pending work. Inodes may see a lot of * write activity after they are unlinked which we must not hold up. */ for (; inodedep != NULL; inodedep = TAILQ_NEXT(inodedep, id_unlinked)) { if ((inodedep->id_state & UNLINKLINKS) != UNLINKLINKS) panic("handle_written_sbdep: Bad inodedep %p (0x%X)", inodedep, inodedep->id_state); if (inodedep->id_state & UNLINKONLIST) break; inodedep->id_state |= DEPCOMPLETE | UNLINKONLIST; } return (0); } /* * Mark an inodedep as unlinked and insert it into the in-memory unlinked list. */ static void unlinked_inodedep(mp, inodedep) struct mount *mp; struct inodedep *inodedep; { struct ufsmount *ump; ump = VFSTOUFS(mp); LOCK_OWNED(ump); if (MOUNTEDSUJ(mp) == 0) return; ump->um_fs->fs_fmod = 1; if (inodedep->id_state & UNLINKED) panic("unlinked_inodedep: %p already unlinked\n", inodedep); inodedep->id_state |= UNLINKED; TAILQ_INSERT_HEAD(&ump->softdep_unlinked, inodedep, id_unlinked); } /* * Remove an inodedep from the unlinked inodedep list. This may require * disk writes if the inode has made it that far. */ static void clear_unlinked_inodedep(inodedep) struct inodedep *inodedep; { struct ufsmount *ump; struct inodedep *idp; struct inodedep *idn; struct fs *fs; struct buf *bp; ino_t ino; ino_t nino; ino_t pino; int error; ump = VFSTOUFS(inodedep->id_list.wk_mp); fs = ump->um_fs; ino = inodedep->id_ino; error = 0; for (;;) { LOCK_OWNED(ump); KASSERT((inodedep->id_state & UNLINKED) != 0, ("clear_unlinked_inodedep: inodedep %p not unlinked", inodedep)); /* * If nothing has yet been written simply remove us from * the in memory list and return. This is the most common * case where handle_workitem_remove() loses the final * reference. */ if ((inodedep->id_state & UNLINKLINKS) == 0) break; /* * If we have a NEXT pointer and no PREV pointer we can simply * clear NEXT's PREV and remove ourselves from the list. Be * careful not to clear PREV if the superblock points at * next as well. */ idn = TAILQ_NEXT(inodedep, id_unlinked); if ((inodedep->id_state & UNLINKLINKS) == UNLINKNEXT) { if (idn && fs->fs_sujfree != idn->id_ino) idn->id_state &= ~UNLINKPREV; break; } /* * Here we have an inodedep which is actually linked into * the list. We must remove it by forcing a write to the * link before us, whether it be the superblock or an inode. * Unfortunately the list may change while we're waiting * on the buf lock for either resource so we must loop until * we lock the right one. If both the superblock and an * inode point to this inode we must clear the inode first * followed by the superblock. */ idp = TAILQ_PREV(inodedep, inodedeplst, id_unlinked); pino = 0; if (idp && (idp->id_state & UNLINKNEXT)) pino = idp->id_ino; FREE_LOCK(ump); if (pino == 0) { bp = getblk(ump->um_devvp, btodb(fs->fs_sblockloc), (int)fs->fs_sbsize, 0, 0, 0); } else { error = bread(ump->um_devvp, fsbtodb(fs, ino_to_fsba(fs, pino)), (int)fs->fs_bsize, NOCRED, &bp); if (error) brelse(bp); } ACQUIRE_LOCK(ump); if (error) break; /* If the list has changed restart the loop. */ idp = TAILQ_PREV(inodedep, inodedeplst, id_unlinked); nino = 0; if (idp && (idp->id_state & UNLINKNEXT)) nino = idp->id_ino; if (nino != pino || (inodedep->id_state & UNLINKPREV) != UNLINKPREV) { FREE_LOCK(ump); brelse(bp); ACQUIRE_LOCK(ump); continue; } nino = 0; idn = TAILQ_NEXT(inodedep, id_unlinked); if (idn) nino = idn->id_ino; /* * Remove us from the in memory list. After this we cannot * access the inodedep. */ KASSERT((inodedep->id_state & UNLINKED) != 0, ("clear_unlinked_inodedep: inodedep %p not unlinked", inodedep)); inodedep->id_state &= ~(UNLINKED | UNLINKLINKS | UNLINKONLIST); TAILQ_REMOVE(&ump->softdep_unlinked, inodedep, id_unlinked); FREE_LOCK(ump); /* * The predecessor's next pointer is manually updated here * so that the NEXT flag is never cleared for an element * that is in the list. */ if (pino == 0) { bcopy((caddr_t)fs, bp->b_data, (u_int)fs->fs_sbsize); ffs_oldfscompat_write((struct fs *)bp->b_data, ump); softdep_setup_sbupdate(ump, (struct fs *)bp->b_data, bp); } else if (fs->fs_magic == FS_UFS1_MAGIC) ((struct ufs1_dinode *)bp->b_data + ino_to_fsbo(fs, pino))->di_freelink = nino; else ((struct ufs2_dinode *)bp->b_data + ino_to_fsbo(fs, pino))->di_freelink = nino; /* * If the bwrite fails we have no recourse to recover. The * filesystem is corrupted already. */ bwrite(bp); ACQUIRE_LOCK(ump); /* * If the superblock pointer still needs to be cleared force * a write here. */ if (fs->fs_sujfree == ino) { FREE_LOCK(ump); bp = getblk(ump->um_devvp, btodb(fs->fs_sblockloc), (int)fs->fs_sbsize, 0, 0, 0); bcopy((caddr_t)fs, bp->b_data, (u_int)fs->fs_sbsize); ffs_oldfscompat_write((struct fs *)bp->b_data, ump); softdep_setup_sbupdate(ump, (struct fs *)bp->b_data, bp); bwrite(bp); ACQUIRE_LOCK(ump); } if (fs->fs_sujfree != ino) return; panic("clear_unlinked_inodedep: Failed to clear free head"); } if (inodedep->id_ino == fs->fs_sujfree) panic("clear_unlinked_inodedep: Freeing head of free list"); inodedep->id_state &= ~(UNLINKED | UNLINKLINKS | UNLINKONLIST); TAILQ_REMOVE(&ump->softdep_unlinked, inodedep, id_unlinked); return; } /* * This workitem decrements the inode's link count. * If the link count reaches zero, the file is removed. */ static int handle_workitem_remove(dirrem, flags) struct dirrem *dirrem; int flags; { struct inodedep *inodedep; struct workhead dotdotwk; struct worklist *wk; struct ufsmount *ump; struct mount *mp; struct vnode *vp; struct inode *ip; ino_t oldinum; if (dirrem->dm_state & ONWORKLIST) panic("handle_workitem_remove: dirrem %p still on worklist", dirrem); oldinum = dirrem->dm_oldinum; mp = dirrem->dm_list.wk_mp; ump = VFSTOUFS(mp); flags |= LK_EXCLUSIVE; if (ffs_vgetf(mp, oldinum, flags, &vp, FFSV_FORCEINSMQ) != 0) return (EBUSY); ip = VTOI(vp); ACQUIRE_LOCK(ump); if ((inodedep_lookup(mp, oldinum, 0, &inodedep)) == 0) panic("handle_workitem_remove: lost inodedep"); if (dirrem->dm_state & ONDEPLIST) LIST_REMOVE(dirrem, dm_inonext); KASSERT(LIST_EMPTY(&dirrem->dm_jremrefhd), ("handle_workitem_remove: Journal entries not written.")); /* * Move all dependencies waiting on the remove to complete * from the dirrem to the inode inowait list to be completed * after the inode has been updated and written to disk. Any * marked MKDIR_PARENT are saved to be completed when the .. ref * is removed. */ LIST_INIT(&dotdotwk); while ((wk = LIST_FIRST(&dirrem->dm_jwork)) != NULL) { WORKLIST_REMOVE(wk); if (wk->wk_state & MKDIR_PARENT) { wk->wk_state &= ~MKDIR_PARENT; WORKLIST_INSERT(&dotdotwk, wk); continue; } WORKLIST_INSERT(&inodedep->id_inowait, wk); } LIST_SWAP(&dirrem->dm_jwork, &dotdotwk, worklist, wk_list); /* * Normal file deletion. */ if ((dirrem->dm_state & RMDIR) == 0) { ip->i_nlink--; DIP_SET(ip, i_nlink, ip->i_nlink); ip->i_flag |= IN_CHANGE; if (ip->i_nlink < ip->i_effnlink) panic("handle_workitem_remove: bad file delta"); if (ip->i_nlink == 0) unlinked_inodedep(mp, inodedep); inodedep->id_nlinkdelta = ip->i_nlink - ip->i_effnlink; KASSERT(LIST_EMPTY(&dirrem->dm_jwork), ("handle_workitem_remove: worklist not empty. %s", TYPENAME(LIST_FIRST(&dirrem->dm_jwork)->wk_type))); WORKITEM_FREE(dirrem, D_DIRREM); FREE_LOCK(ump); goto out; } /* * Directory deletion. Decrement reference count for both the * just deleted parent directory entry and the reference for ".". * Arrange to have the reference count on the parent decremented * to account for the loss of "..". */ ip->i_nlink -= 2; DIP_SET(ip, i_nlink, ip->i_nlink); ip->i_flag |= IN_CHANGE; if (ip->i_nlink < ip->i_effnlink) panic("handle_workitem_remove: bad dir delta"); if (ip->i_nlink == 0) unlinked_inodedep(mp, inodedep); inodedep->id_nlinkdelta = ip->i_nlink - ip->i_effnlink; /* * Rename a directory to a new parent. Since, we are both deleting * and creating a new directory entry, the link count on the new * directory should not change. Thus we skip the followup dirrem. */ if (dirrem->dm_state & DIRCHG) { KASSERT(LIST_EMPTY(&dirrem->dm_jwork), ("handle_workitem_remove: DIRCHG and worklist not empty.")); WORKITEM_FREE(dirrem, D_DIRREM); FREE_LOCK(ump); goto out; } dirrem->dm_state = ONDEPLIST; dirrem->dm_oldinum = dirrem->dm_dirinum; /* * Place the dirrem on the parent's diremhd list. */ if (inodedep_lookup(mp, dirrem->dm_oldinum, 0, &inodedep) == 0) panic("handle_workitem_remove: lost dir inodedep"); LIST_INSERT_HEAD(&inodedep->id_dirremhd, dirrem, dm_inonext); /* * If the allocated inode has never been written to disk, then * the on-disk inode is zero'ed and we can remove the file * immediately. When journaling if the inode has been marked * unlinked and not DEPCOMPLETE we know it can never be written. */ inodedep_lookup(mp, oldinum, 0, &inodedep); if (inodedep == NULL || (inodedep->id_state & (DEPCOMPLETE | UNLINKED)) == UNLINKED || check_inode_unwritten(inodedep)) { FREE_LOCK(ump); vput(vp); return handle_workitem_remove(dirrem, flags); } WORKLIST_INSERT(&inodedep->id_inowait, &dirrem->dm_list); FREE_LOCK(ump); ip->i_flag |= IN_CHANGE; out: ffs_update(vp, 0); vput(vp); return (0); } /* * Inode de-allocation dependencies. * * When an inode's link count is reduced to zero, it can be de-allocated. We * found it convenient to postpone de-allocation until after the inode is * written to disk with its new link count (zero). At this point, all of the * on-disk inode's block pointers are nullified and, with careful dependency * list ordering, all dependencies related to the inode will be satisfied and * the corresponding dependency structures de-allocated. So, if/when the * inode is reused, there will be no mixing of old dependencies with new * ones. This artificial dependency is set up by the block de-allocation * procedure above (softdep_setup_freeblocks) and completed by the * following procedure. */ static void handle_workitem_freefile(freefile) struct freefile *freefile; { struct workhead wkhd; struct fs *fs; struct inodedep *idp; struct ufsmount *ump; int error; ump = VFSTOUFS(freefile->fx_list.wk_mp); fs = ump->um_fs; #ifdef DEBUG ACQUIRE_LOCK(ump); error = inodedep_lookup(UFSTOVFS(ump), freefile->fx_oldinum, 0, &idp); FREE_LOCK(ump); if (error) panic("handle_workitem_freefile: inodedep %p survived", idp); #endif UFS_LOCK(ump); fs->fs_pendinginodes -= 1; UFS_UNLOCK(ump); LIST_INIT(&wkhd); LIST_SWAP(&freefile->fx_jwork, &wkhd, worklist, wk_list); if ((error = ffs_freefile(ump, fs, freefile->fx_devvp, freefile->fx_oldinum, freefile->fx_mode, &wkhd)) != 0) softdep_error("handle_workitem_freefile", error); ACQUIRE_LOCK(ump); WORKITEM_FREE(freefile, D_FREEFILE); FREE_LOCK(ump); } /* * Helper function which unlinks marker element from work list and returns * the next element on the list. */ static __inline struct worklist * markernext(struct worklist *marker) { struct worklist *next; next = LIST_NEXT(marker, wk_list); LIST_REMOVE(marker, wk_list); return next; } /* * Disk writes. * * The dependency structures constructed above are most actively used when file * system blocks are written to disk. No constraints are placed on when a * block can be written, but unsatisfied update dependencies are made safe by * modifying (or replacing) the source memory for the duration of the disk * write. When the disk write completes, the memory block is again brought * up-to-date. * * In-core inode structure reclamation. * * Because there are a finite number of "in-core" inode structures, they are * reused regularly. By transferring all inode-related dependencies to the * in-memory inode block and indexing them separately (via "inodedep"s), we * can allow "in-core" inode structures to be reused at any time and avoid * any increase in contention. * * Called just before entering the device driver to initiate a new disk I/O. * The buffer must be locked, thus, no I/O completion operations can occur * while we are manipulating its associated dependencies. */ static void softdep_disk_io_initiation(bp) struct buf *bp; /* structure describing disk write to occur */ { struct worklist *wk; struct worklist marker; struct inodedep *inodedep; struct freeblks *freeblks; struct jblkdep *jblkdep; struct newblk *newblk; struct ufsmount *ump; /* * We only care about write operations. There should never * be dependencies for reads. */ if (bp->b_iocmd != BIO_WRITE) panic("softdep_disk_io_initiation: not write"); if (bp->b_vflags & BV_BKGRDINPROG) panic("softdep_disk_io_initiation: Writing buffer with " "background write in progress: %p", bp); if ((wk = LIST_FIRST(&bp->b_dep)) == NULL) return; ump = VFSTOUFS(wk->wk_mp); marker.wk_type = D_LAST + 1; /* Not a normal workitem */ PHOLD(curproc); /* Don't swap out kernel stack */ ACQUIRE_LOCK(ump); /* * Do any necessary pre-I/O processing. */ for (wk = LIST_FIRST(&bp->b_dep); wk != NULL; wk = markernext(&marker)) { LIST_INSERT_AFTER(wk, &marker, wk_list); switch (wk->wk_type) { case D_PAGEDEP: initiate_write_filepage(WK_PAGEDEP(wk), bp); continue; case D_INODEDEP: inodedep = WK_INODEDEP(wk); if (inodedep->id_fs->fs_magic == FS_UFS1_MAGIC) initiate_write_inodeblock_ufs1(inodedep, bp); else initiate_write_inodeblock_ufs2(inodedep, bp); continue; case D_INDIRDEP: initiate_write_indirdep(WK_INDIRDEP(wk), bp); continue; case D_BMSAFEMAP: initiate_write_bmsafemap(WK_BMSAFEMAP(wk), bp); continue; case D_JSEG: WK_JSEG(wk)->js_buf = NULL; continue; case D_FREEBLKS: freeblks = WK_FREEBLKS(wk); jblkdep = LIST_FIRST(&freeblks->fb_jblkdephd); /* * We have to wait for the freeblks to be journaled * before we can write an inodeblock with updated * pointers. Be careful to arrange the marker so * we revisit the freeblks if it's not removed by * the first jwait(). */ if (jblkdep != NULL) { LIST_REMOVE(&marker, wk_list); LIST_INSERT_BEFORE(wk, &marker, wk_list); jwait(&jblkdep->jb_list, MNT_WAIT); } continue; case D_ALLOCDIRECT: case D_ALLOCINDIR: /* * We have to wait for the jnewblk to be journaled * before we can write to a block if the contents * may be confused with an earlier file's indirect * at recovery time. Handle the marker as described * above. */ newblk = WK_NEWBLK(wk); if (newblk->nb_jnewblk != NULL && indirblk_lookup(newblk->nb_list.wk_mp, newblk->nb_newblkno)) { LIST_REMOVE(&marker, wk_list); LIST_INSERT_BEFORE(wk, &marker, wk_list); jwait(&newblk->nb_jnewblk->jn_list, MNT_WAIT); } continue; case D_SBDEP: initiate_write_sbdep(WK_SBDEP(wk)); continue; case D_MKDIR: case D_FREEWORK: case D_FREEDEP: case D_JSEGDEP: continue; default: panic("handle_disk_io_initiation: Unexpected type %s", TYPENAME(wk->wk_type)); /* NOTREACHED */ } } FREE_LOCK(ump); PRELE(curproc); /* Allow swapout of kernel stack */ } /* * Called from within the procedure above to deal with unsatisfied * allocation dependencies in a directory. The buffer must be locked, * thus, no I/O completion operations can occur while we are * manipulating its associated dependencies. */ static void initiate_write_filepage(pagedep, bp) struct pagedep *pagedep; struct buf *bp; { struct jremref *jremref; struct jmvref *jmvref; struct dirrem *dirrem; struct diradd *dap; struct direct *ep; int i; if (pagedep->pd_state & IOSTARTED) { /* * This can only happen if there is a driver that does not * understand chaining. Here biodone will reissue the call * to strategy for the incomplete buffers. */ printf("initiate_write_filepage: already started\n"); return; } pagedep->pd_state |= IOSTARTED; /* * Wait for all journal remove dependencies to hit the disk. * We can not allow any potentially conflicting directory adds * to be visible before removes and rollback is too difficult. * The per-filesystem lock may be dropped and re-acquired, however * we hold the buf locked so the dependency can not go away. */ LIST_FOREACH(dirrem, &pagedep->pd_dirremhd, dm_next) while ((jremref = LIST_FIRST(&dirrem->dm_jremrefhd)) != NULL) jwait(&jremref->jr_list, MNT_WAIT); while ((jmvref = LIST_FIRST(&pagedep->pd_jmvrefhd)) != NULL) jwait(&jmvref->jm_list, MNT_WAIT); for (i = 0; i < DAHASHSZ; i++) { LIST_FOREACH(dap, &pagedep->pd_diraddhd[i], da_pdlist) { ep = (struct direct *) ((char *)bp->b_data + dap->da_offset); if (ep->d_ino != dap->da_newinum) panic("%s: dir inum %ju != new %ju", "initiate_write_filepage", (uintmax_t)ep->d_ino, (uintmax_t)dap->da_newinum); if (dap->da_state & DIRCHG) ep->d_ino = dap->da_previous->dm_oldinum; else ep->d_ino = 0; dap->da_state &= ~ATTACHED; dap->da_state |= UNDONE; } } } /* * Version of initiate_write_inodeblock that handles UFS1 dinodes. * Note that any bug fixes made to this routine must be done in the * version found below. * * Called from within the procedure above to deal with unsatisfied * allocation dependencies in an inodeblock. The buffer must be * locked, thus, no I/O completion operations can occur while we * are manipulating its associated dependencies. */ static void initiate_write_inodeblock_ufs1(inodedep, bp) struct inodedep *inodedep; struct buf *bp; /* The inode block */ { struct allocdirect *adp, *lastadp; struct ufs1_dinode *dp; struct ufs1_dinode *sip; struct inoref *inoref; struct ufsmount *ump; struct fs *fs; ufs_lbn_t i; #ifdef INVARIANTS ufs_lbn_t prevlbn = 0; #endif int deplist; if (inodedep->id_state & IOSTARTED) panic("initiate_write_inodeblock_ufs1: already started"); inodedep->id_state |= IOSTARTED; fs = inodedep->id_fs; ump = VFSTOUFS(inodedep->id_list.wk_mp); LOCK_OWNED(ump); dp = (struct ufs1_dinode *)bp->b_data + ino_to_fsbo(fs, inodedep->id_ino); /* * If we're on the unlinked list but have not yet written our * next pointer initialize it here. */ if ((inodedep->id_state & (UNLINKED | UNLINKNEXT)) == UNLINKED) { struct inodedep *inon; inon = TAILQ_NEXT(inodedep, id_unlinked); dp->di_freelink = inon ? inon->id_ino : 0; } /* * If the bitmap is not yet written, then the allocated * inode cannot be written to disk. */ if ((inodedep->id_state & DEPCOMPLETE) == 0) { if (inodedep->id_savedino1 != NULL) panic("initiate_write_inodeblock_ufs1: I/O underway"); FREE_LOCK(ump); sip = malloc(sizeof(struct ufs1_dinode), M_SAVEDINO, M_SOFTDEP_FLAGS); ACQUIRE_LOCK(ump); inodedep->id_savedino1 = sip; *inodedep->id_savedino1 = *dp; bzero((caddr_t)dp, sizeof(struct ufs1_dinode)); dp->di_gen = inodedep->id_savedino1->di_gen; dp->di_freelink = inodedep->id_savedino1->di_freelink; return; } /* * If no dependencies, then there is nothing to roll back. */ inodedep->id_savedsize = dp->di_size; inodedep->id_savedextsize = 0; inodedep->id_savednlink = dp->di_nlink; if (TAILQ_EMPTY(&inodedep->id_inoupdt) && TAILQ_EMPTY(&inodedep->id_inoreflst)) return; /* * Revert the link count to that of the first unwritten journal entry. */ inoref = TAILQ_FIRST(&inodedep->id_inoreflst); if (inoref) dp->di_nlink = inoref->if_nlink; /* * Set the dependencies to busy. */ for (deplist = 0, adp = TAILQ_FIRST(&inodedep->id_inoupdt); adp; adp = TAILQ_NEXT(adp, ad_next)) { #ifdef INVARIANTS if (deplist != 0 && prevlbn >= adp->ad_offset) panic("softdep_write_inodeblock: lbn order"); prevlbn = adp->ad_offset; if (adp->ad_offset < NDADDR && dp->di_db[adp->ad_offset] != adp->ad_newblkno) panic("%s: direct pointer #%jd mismatch %d != %jd", "softdep_write_inodeblock", (intmax_t)adp->ad_offset, dp->di_db[adp->ad_offset], (intmax_t)adp->ad_newblkno); if (adp->ad_offset >= NDADDR && dp->di_ib[adp->ad_offset - NDADDR] != adp->ad_newblkno) panic("%s: indirect pointer #%jd mismatch %d != %jd", "softdep_write_inodeblock", (intmax_t)adp->ad_offset - NDADDR, dp->di_ib[adp->ad_offset - NDADDR], (intmax_t)adp->ad_newblkno); deplist |= 1 << adp->ad_offset; if ((adp->ad_state & ATTACHED) == 0) panic("softdep_write_inodeblock: Unknown state 0x%x", adp->ad_state); #endif /* INVARIANTS */ adp->ad_state &= ~ATTACHED; adp->ad_state |= UNDONE; } /* * The on-disk inode cannot claim to be any larger than the last * fragment that has been written. Otherwise, the on-disk inode * might have fragments that were not the last block in the file * which would corrupt the filesystem. */ for (lastadp = NULL, adp = TAILQ_FIRST(&inodedep->id_inoupdt); adp; lastadp = adp, adp = TAILQ_NEXT(adp, ad_next)) { if (adp->ad_offset >= NDADDR) break; dp->di_db[adp->ad_offset] = adp->ad_oldblkno; /* keep going until hitting a rollback to a frag */ if (adp->ad_oldsize == 0 || adp->ad_oldsize == fs->fs_bsize) continue; dp->di_size = fs->fs_bsize * adp->ad_offset + adp->ad_oldsize; for (i = adp->ad_offset + 1; i < NDADDR; i++) { #ifdef INVARIANTS if (dp->di_db[i] != 0 && (deplist & (1 << i)) == 0) panic("softdep_write_inodeblock: lost dep1"); #endif /* INVARIANTS */ dp->di_db[i] = 0; } for (i = 0; i < NIADDR; i++) { #ifdef INVARIANTS if (dp->di_ib[i] != 0 && (deplist & ((1 << NDADDR) << i)) == 0) panic("softdep_write_inodeblock: lost dep2"); #endif /* INVARIANTS */ dp->di_ib[i] = 0; } return; } /* * If we have zero'ed out the last allocated block of the file, * roll back the size to the last currently allocated block. * We know that this last allocated block is a full-sized as * we already checked for fragments in the loop above. */ if (lastadp != NULL && dp->di_size <= (lastadp->ad_offset + 1) * fs->fs_bsize) { for (i = lastadp->ad_offset; i >= 0; i--) if (dp->di_db[i] != 0) break; dp->di_size = (i + 1) * fs->fs_bsize; } /* * The only dependencies are for indirect blocks. * * The file size for indirect block additions is not guaranteed. * Such a guarantee would be non-trivial to achieve. The conventional * synchronous write implementation also does not make this guarantee. * Fsck should catch and fix discrepancies. Arguably, the file size * can be over-estimated without destroying integrity when the file * moves into the indirect blocks (i.e., is large). If we want to * postpone fsck, we are stuck with this argument. */ for (; adp; adp = TAILQ_NEXT(adp, ad_next)) dp->di_ib[adp->ad_offset - NDADDR] = 0; } /* * Version of initiate_write_inodeblock that handles UFS2 dinodes. * Note that any bug fixes made to this routine must be done in the * version found above. * * Called from within the procedure above to deal with unsatisfied * allocation dependencies in an inodeblock. The buffer must be * locked, thus, no I/O completion operations can occur while we * are manipulating its associated dependencies. */ static void initiate_write_inodeblock_ufs2(inodedep, bp) struct inodedep *inodedep; struct buf *bp; /* The inode block */ { struct allocdirect *adp, *lastadp; struct ufs2_dinode *dp; struct ufs2_dinode *sip; struct inoref *inoref; struct ufsmount *ump; struct fs *fs; ufs_lbn_t i; #ifdef INVARIANTS ufs_lbn_t prevlbn = 0; #endif int deplist; if (inodedep->id_state & IOSTARTED) panic("initiate_write_inodeblock_ufs2: already started"); inodedep->id_state |= IOSTARTED; fs = inodedep->id_fs; ump = VFSTOUFS(inodedep->id_list.wk_mp); LOCK_OWNED(ump); dp = (struct ufs2_dinode *)bp->b_data + ino_to_fsbo(fs, inodedep->id_ino); /* * If we're on the unlinked list but have not yet written our * next pointer initialize it here. */ if ((inodedep->id_state & (UNLINKED | UNLINKNEXT)) == UNLINKED) { struct inodedep *inon; inon = TAILQ_NEXT(inodedep, id_unlinked); dp->di_freelink = inon ? inon->id_ino : 0; } /* * If the bitmap is not yet written, then the allocated * inode cannot be written to disk. */ if ((inodedep->id_state & DEPCOMPLETE) == 0) { if (inodedep->id_savedino2 != NULL) panic("initiate_write_inodeblock_ufs2: I/O underway"); FREE_LOCK(ump); sip = malloc(sizeof(struct ufs2_dinode), M_SAVEDINO, M_SOFTDEP_FLAGS); ACQUIRE_LOCK(ump); inodedep->id_savedino2 = sip; *inodedep->id_savedino2 = *dp; bzero((caddr_t)dp, sizeof(struct ufs2_dinode)); dp->di_gen = inodedep->id_savedino2->di_gen; dp->di_freelink = inodedep->id_savedino2->di_freelink; return; } /* * If no dependencies, then there is nothing to roll back. */ inodedep->id_savedsize = dp->di_size; inodedep->id_savedextsize = dp->di_extsize; inodedep->id_savednlink = dp->di_nlink; if (TAILQ_EMPTY(&inodedep->id_inoupdt) && TAILQ_EMPTY(&inodedep->id_extupdt) && TAILQ_EMPTY(&inodedep->id_inoreflst)) return; /* * Revert the link count to that of the first unwritten journal entry. */ inoref = TAILQ_FIRST(&inodedep->id_inoreflst); if (inoref) dp->di_nlink = inoref->if_nlink; /* * Set the ext data dependencies to busy. */ for (deplist = 0, adp = TAILQ_FIRST(&inodedep->id_extupdt); adp; adp = TAILQ_NEXT(adp, ad_next)) { #ifdef INVARIANTS if (deplist != 0 && prevlbn >= adp->ad_offset) panic("softdep_write_inodeblock: lbn order"); prevlbn = adp->ad_offset; if (dp->di_extb[adp->ad_offset] != adp->ad_newblkno) panic("%s: direct pointer #%jd mismatch %jd != %jd", "softdep_write_inodeblock", (intmax_t)adp->ad_offset, (intmax_t)dp->di_extb[adp->ad_offset], (intmax_t)adp->ad_newblkno); deplist |= 1 << adp->ad_offset; if ((adp->ad_state & ATTACHED) == 0) panic("softdep_write_inodeblock: Unknown state 0x%x", adp->ad_state); #endif /* INVARIANTS */ adp->ad_state &= ~ATTACHED; adp->ad_state |= UNDONE; } /* * The on-disk inode cannot claim to be any larger than the last * fragment that has been written. Otherwise, the on-disk inode * might have fragments that were not the last block in the ext * data which would corrupt the filesystem. */ for (lastadp = NULL, adp = TAILQ_FIRST(&inodedep->id_extupdt); adp; lastadp = adp, adp = TAILQ_NEXT(adp, ad_next)) { dp->di_extb[adp->ad_offset] = adp->ad_oldblkno; /* keep going until hitting a rollback to a frag */ if (adp->ad_oldsize == 0 || adp->ad_oldsize == fs->fs_bsize) continue; dp->di_extsize = fs->fs_bsize * adp->ad_offset + adp->ad_oldsize; for (i = adp->ad_offset + 1; i < NXADDR; i++) { #ifdef INVARIANTS if (dp->di_extb[i] != 0 && (deplist & (1 << i)) == 0) panic("softdep_write_inodeblock: lost dep1"); #endif /* INVARIANTS */ dp->di_extb[i] = 0; } lastadp = NULL; break; } /* * If we have zero'ed out the last allocated block of the ext * data, roll back the size to the last currently allocated block. * We know that this last allocated block is a full-sized as * we already checked for fragments in the loop above. */ if (lastadp != NULL && dp->di_extsize <= (lastadp->ad_offset + 1) * fs->fs_bsize) { for (i = lastadp->ad_offset; i >= 0; i--) if (dp->di_extb[i] != 0) break; dp->di_extsize = (i + 1) * fs->fs_bsize; } /* * Set the file data dependencies to busy. */ for (deplist = 0, adp = TAILQ_FIRST(&inodedep->id_inoupdt); adp; adp = TAILQ_NEXT(adp, ad_next)) { #ifdef INVARIANTS if (deplist != 0 && prevlbn >= adp->ad_offset) panic("softdep_write_inodeblock: lbn order"); if ((adp->ad_state & ATTACHED) == 0) panic("inodedep %p and adp %p not attached", inodedep, adp); prevlbn = adp->ad_offset; if (adp->ad_offset < NDADDR && dp->di_db[adp->ad_offset] != adp->ad_newblkno) panic("%s: direct pointer #%jd mismatch %jd != %jd", "softdep_write_inodeblock", (intmax_t)adp->ad_offset, (intmax_t)dp->di_db[adp->ad_offset], (intmax_t)adp->ad_newblkno); if (adp->ad_offset >= NDADDR && dp->di_ib[adp->ad_offset - NDADDR] != adp->ad_newblkno) panic("%s indirect pointer #%jd mismatch %jd != %jd", "softdep_write_inodeblock:", (intmax_t)adp->ad_offset - NDADDR, (intmax_t)dp->di_ib[adp->ad_offset - NDADDR], (intmax_t)adp->ad_newblkno); deplist |= 1 << adp->ad_offset; if ((adp->ad_state & ATTACHED) == 0) panic("softdep_write_inodeblock: Unknown state 0x%x", adp->ad_state); #endif /* INVARIANTS */ adp->ad_state &= ~ATTACHED; adp->ad_state |= UNDONE; } /* * The on-disk inode cannot claim to be any larger than the last * fragment that has been written. Otherwise, the on-disk inode * might have fragments that were not the last block in the file * which would corrupt the filesystem. */ for (lastadp = NULL, adp = TAILQ_FIRST(&inodedep->id_inoupdt); adp; lastadp = adp, adp = TAILQ_NEXT(adp, ad_next)) { if (adp->ad_offset >= NDADDR) break; dp->di_db[adp->ad_offset] = adp->ad_oldblkno; /* keep going until hitting a rollback to a frag */ if (adp->ad_oldsize == 0 || adp->ad_oldsize == fs->fs_bsize) continue; dp->di_size = fs->fs_bsize * adp->ad_offset + adp->ad_oldsize; for (i = adp->ad_offset + 1; i < NDADDR; i++) { #ifdef INVARIANTS if (dp->di_db[i] != 0 && (deplist & (1 << i)) == 0) panic("softdep_write_inodeblock: lost dep2"); #endif /* INVARIANTS */ dp->di_db[i] = 0; } for (i = 0; i < NIADDR; i++) { #ifdef INVARIANTS if (dp->di_ib[i] != 0 && (deplist & ((1 << NDADDR) << i)) == 0) panic("softdep_write_inodeblock: lost dep3"); #endif /* INVARIANTS */ dp->di_ib[i] = 0; } return; } /* * If we have zero'ed out the last allocated block of the file, * roll back the size to the last currently allocated block. * We know that this last allocated block is a full-sized as * we already checked for fragments in the loop above. */ if (lastadp != NULL && dp->di_size <= (lastadp->ad_offset + 1) * fs->fs_bsize) { for (i = lastadp->ad_offset; i >= 0; i--) if (dp->di_db[i] != 0) break; dp->di_size = (i + 1) * fs->fs_bsize; } /* * The only dependencies are for indirect blocks. * * The file size for indirect block additions is not guaranteed. * Such a guarantee would be non-trivial to achieve. The conventional * synchronous write implementation also does not make this guarantee. * Fsck should catch and fix discrepancies. Arguably, the file size * can be over-estimated without destroying integrity when the file * moves into the indirect blocks (i.e., is large). If we want to * postpone fsck, we are stuck with this argument. */ for (; adp; adp = TAILQ_NEXT(adp, ad_next)) dp->di_ib[adp->ad_offset - NDADDR] = 0; } /* * Cancel an indirdep as a result of truncation. Release all of the * children allocindirs and place their journal work on the appropriate * list. */ static void cancel_indirdep(indirdep, bp, freeblks) struct indirdep *indirdep; struct buf *bp; struct freeblks *freeblks; { struct allocindir *aip; /* * None of the indirect pointers will ever be visible, * so they can simply be tossed. GOINGAWAY ensures * that allocated pointers will be saved in the buffer * cache until they are freed. Note that they will * only be able to be found by their physical address * since the inode mapping the logical address will * be gone. The save buffer used for the safe copy * was allocated in setup_allocindir_phase2 using * the physical address so it could be used for this * purpose. Hence we swap the safe copy with the real * copy, allowing the safe copy to be freed and holding * on to the real copy for later use in indir_trunc. */ if (indirdep->ir_state & GOINGAWAY) panic("cancel_indirdep: already gone"); if ((indirdep->ir_state & DEPCOMPLETE) == 0) { indirdep->ir_state |= DEPCOMPLETE; LIST_REMOVE(indirdep, ir_next); } indirdep->ir_state |= GOINGAWAY; /* * Pass in bp for blocks still have journal writes * pending so we can cancel them on their own. */ while ((aip = LIST_FIRST(&indirdep->ir_deplisthd)) != 0) cancel_allocindir(aip, bp, freeblks, 0); while ((aip = LIST_FIRST(&indirdep->ir_donehd)) != 0) cancel_allocindir(aip, NULL, freeblks, 0); while ((aip = LIST_FIRST(&indirdep->ir_writehd)) != 0) cancel_allocindir(aip, NULL, freeblks, 0); while ((aip = LIST_FIRST(&indirdep->ir_completehd)) != 0) cancel_allocindir(aip, NULL, freeblks, 0); /* * If there are pending partial truncations we need to keep the * old block copy around until they complete. This is because * the current b_data is not a perfect superset of the available * blocks. */ if (TAILQ_EMPTY(&indirdep->ir_trunc)) bcopy(bp->b_data, indirdep->ir_savebp->b_data, bp->b_bcount); else bcopy(bp->b_data, indirdep->ir_saveddata, bp->b_bcount); WORKLIST_REMOVE(&indirdep->ir_list); WORKLIST_INSERT(&indirdep->ir_savebp->b_dep, &indirdep->ir_list); indirdep->ir_bp = NULL; indirdep->ir_freeblks = freeblks; } /* * Free an indirdep once it no longer has new pointers to track. */ static void free_indirdep(indirdep) struct indirdep *indirdep; { KASSERT(TAILQ_EMPTY(&indirdep->ir_trunc), ("free_indirdep: Indir trunc list not empty.")); KASSERT(LIST_EMPTY(&indirdep->ir_completehd), ("free_indirdep: Complete head not empty.")); KASSERT(LIST_EMPTY(&indirdep->ir_writehd), ("free_indirdep: write head not empty.")); KASSERT(LIST_EMPTY(&indirdep->ir_donehd), ("free_indirdep: done head not empty.")); KASSERT(LIST_EMPTY(&indirdep->ir_deplisthd), ("free_indirdep: deplist head not empty.")); KASSERT((indirdep->ir_state & DEPCOMPLETE), ("free_indirdep: %p still on newblk list.", indirdep)); KASSERT(indirdep->ir_saveddata == NULL, ("free_indirdep: %p still has saved data.", indirdep)); if (indirdep->ir_state & ONWORKLIST) WORKLIST_REMOVE(&indirdep->ir_list); WORKITEM_FREE(indirdep, D_INDIRDEP); } /* * Called before a write to an indirdep. This routine is responsible for * rolling back pointers to a safe state which includes only those * allocindirs which have been completed. */ static void initiate_write_indirdep(indirdep, bp) struct indirdep *indirdep; struct buf *bp; { struct ufsmount *ump; indirdep->ir_state |= IOSTARTED; if (indirdep->ir_state & GOINGAWAY) panic("disk_io_initiation: indirdep gone"); /* * If there are no remaining dependencies, this will be writing * the real pointers. */ if (LIST_EMPTY(&indirdep->ir_deplisthd) && TAILQ_EMPTY(&indirdep->ir_trunc)) return; /* * Replace up-to-date version with safe version. */ if (indirdep->ir_saveddata == NULL) { ump = VFSTOUFS(indirdep->ir_list.wk_mp); LOCK_OWNED(ump); FREE_LOCK(ump); indirdep->ir_saveddata = malloc(bp->b_bcount, M_INDIRDEP, M_SOFTDEP_FLAGS); ACQUIRE_LOCK(ump); } indirdep->ir_state &= ~ATTACHED; indirdep->ir_state |= UNDONE; bcopy(bp->b_data, indirdep->ir_saveddata, bp->b_bcount); bcopy(indirdep->ir_savebp->b_data, bp->b_data, bp->b_bcount); } /* * Called when an inode has been cleared in a cg bitmap. This finally * eliminates any canceled jaddrefs */ void softdep_setup_inofree(mp, bp, ino, wkhd) struct mount *mp; struct buf *bp; ino_t ino; struct workhead *wkhd; { struct worklist *wk, *wkn; struct inodedep *inodedep; struct ufsmount *ump; uint8_t *inosused; struct cg *cgp; struct fs *fs; KASSERT(MOUNTEDSOFTDEP(mp) != 0, ("softdep_setup_inofree called on non-softdep filesystem")); ump = VFSTOUFS(mp); ACQUIRE_LOCK(ump); fs = ump->um_fs; cgp = (struct cg *)bp->b_data; inosused = cg_inosused(cgp); if (isset(inosused, ino % fs->fs_ipg)) panic("softdep_setup_inofree: inode %ju not freed.", (uintmax_t)ino); if (inodedep_lookup(mp, ino, 0, &inodedep)) panic("softdep_setup_inofree: ino %ju has existing inodedep %p", (uintmax_t)ino, inodedep); if (wkhd) { LIST_FOREACH_SAFE(wk, wkhd, wk_list, wkn) { if (wk->wk_type != D_JADDREF) continue; WORKLIST_REMOVE(wk); /* * We can free immediately even if the jaddref * isn't attached in a background write as now * the bitmaps are reconciled. */ wk->wk_state |= COMPLETE | ATTACHED; free_jaddref(WK_JADDREF(wk)); } jwork_move(&bp->b_dep, wkhd); } FREE_LOCK(ump); } /* * Called via ffs_blkfree() after a set of frags has been cleared from a cg * map. Any dependencies waiting for the write to clear are added to the * buf's list and any jnewblks that are being canceled are discarded * immediately. */ void softdep_setup_blkfree(mp, bp, blkno, frags, wkhd) struct mount *mp; struct buf *bp; ufs2_daddr_t blkno; int frags; struct workhead *wkhd; { struct bmsafemap *bmsafemap; struct jnewblk *jnewblk; struct ufsmount *ump; struct worklist *wk; struct fs *fs; #ifdef SUJ_DEBUG uint8_t *blksfree; struct cg *cgp; ufs2_daddr_t jstart; ufs2_daddr_t jend; ufs2_daddr_t end; long bno; int i; #endif CTR3(KTR_SUJ, "softdep_setup_blkfree: blkno %jd frags %d wk head %p", blkno, frags, wkhd); ump = VFSTOUFS(mp); KASSERT(MOUNTEDSOFTDEP(UFSTOVFS(ump)) != 0, ("softdep_setup_blkfree called on non-softdep filesystem")); ACQUIRE_LOCK(ump); /* Lookup the bmsafemap so we track when it is dirty. */ fs = ump->um_fs; bmsafemap = bmsafemap_lookup(mp, bp, dtog(fs, blkno), NULL); /* * Detach any jnewblks which have been canceled. They must linger * until the bitmap is cleared again by ffs_blkfree() to prevent * an unjournaled allocation from hitting the disk. */ if (wkhd) { while ((wk = LIST_FIRST(wkhd)) != NULL) { CTR2(KTR_SUJ, "softdep_setup_blkfree: blkno %jd wk type %d", blkno, wk->wk_type); WORKLIST_REMOVE(wk); if (wk->wk_type != D_JNEWBLK) { WORKLIST_INSERT(&bmsafemap->sm_freehd, wk); continue; } jnewblk = WK_JNEWBLK(wk); KASSERT(jnewblk->jn_state & GOINGAWAY, ("softdep_setup_blkfree: jnewblk not canceled.")); #ifdef SUJ_DEBUG /* * Assert that this block is free in the bitmap * before we discard the jnewblk. */ cgp = (struct cg *)bp->b_data; blksfree = cg_blksfree(cgp); bno = dtogd(fs, jnewblk->jn_blkno); for (i = jnewblk->jn_oldfrags; i < jnewblk->jn_frags; i++) { if (isset(blksfree, bno + i)) continue; panic("softdep_setup_blkfree: not free"); } #endif /* * Even if it's not attached we can free immediately * as the new bitmap is correct. */ wk->wk_state |= COMPLETE | ATTACHED; free_jnewblk(jnewblk); } } #ifdef SUJ_DEBUG /* * Assert that we are not freeing a block which has an outstanding * allocation dependency. */ fs = VFSTOUFS(mp)->um_fs; bmsafemap = bmsafemap_lookup(mp, bp, dtog(fs, blkno), NULL); end = blkno + frags; LIST_FOREACH(jnewblk, &bmsafemap->sm_jnewblkhd, jn_deps) { /* * Don't match against blocks that will be freed when the * background write is done. */ if ((jnewblk->jn_state & (ATTACHED | COMPLETE | DEPCOMPLETE)) == (COMPLETE | DEPCOMPLETE)) continue; jstart = jnewblk->jn_blkno + jnewblk->jn_oldfrags; jend = jnewblk->jn_blkno + jnewblk->jn_frags; if ((blkno >= jstart && blkno < jend) || (end > jstart && end <= jend)) { printf("state 0x%X %jd - %d %d dep %p\n", jnewblk->jn_state, jnewblk->jn_blkno, jnewblk->jn_oldfrags, jnewblk->jn_frags, jnewblk->jn_dep); panic("softdep_setup_blkfree: " "%jd-%jd(%d) overlaps with %jd-%jd", blkno, end, frags, jstart, jend); } } #endif FREE_LOCK(ump); } /* * Revert a block allocation when the journal record that describes it * is not yet written. */ static int jnewblk_rollback(jnewblk, fs, cgp, blksfree) struct jnewblk *jnewblk; struct fs *fs; struct cg *cgp; uint8_t *blksfree; { ufs1_daddr_t fragno; long cgbno, bbase; int frags, blk; int i; frags = 0; cgbno = dtogd(fs, jnewblk->jn_blkno); /* * We have to test which frags need to be rolled back. We may * be operating on a stale copy when doing background writes. */ for (i = jnewblk->jn_oldfrags; i < jnewblk->jn_frags; i++) if (isclr(blksfree, cgbno + i)) frags++; if (frags == 0) return (0); /* * This is mostly ffs_blkfree() sans some validation and * superblock updates. */ if (frags == fs->fs_frag) { fragno = fragstoblks(fs, cgbno); ffs_setblock(fs, blksfree, fragno); ffs_clusteracct(fs, cgp, fragno, 1); cgp->cg_cs.cs_nbfree++; } else { cgbno += jnewblk->jn_oldfrags; bbase = cgbno - fragnum(fs, cgbno); /* Decrement the old frags. */ blk = blkmap(fs, blksfree, bbase); ffs_fragacct(fs, blk, cgp->cg_frsum, -1); /* Deallocate the fragment */ for (i = 0; i < frags; i++) setbit(blksfree, cgbno + i); cgp->cg_cs.cs_nffree += frags; /* Add back in counts associated with the new frags */ blk = blkmap(fs, blksfree, bbase); ffs_fragacct(fs, blk, cgp->cg_frsum, 1); /* If a complete block has been reassembled, account for it. */ fragno = fragstoblks(fs, bbase); if (ffs_isblock(fs, blksfree, fragno)) { cgp->cg_cs.cs_nffree -= fs->fs_frag; ffs_clusteracct(fs, cgp, fragno, 1); cgp->cg_cs.cs_nbfree++; } } stat_jnewblk++; jnewblk->jn_state &= ~ATTACHED; jnewblk->jn_state |= UNDONE; return (frags); } static void initiate_write_bmsafemap(bmsafemap, bp) struct bmsafemap *bmsafemap; struct buf *bp; /* The cg block. */ { struct jaddref *jaddref; struct jnewblk *jnewblk; uint8_t *inosused; uint8_t *blksfree; struct cg *cgp; struct fs *fs; ino_t ino; if (bmsafemap->sm_state & IOSTARTED) return; bmsafemap->sm_state |= IOSTARTED; /* * Clear any inode allocations which are pending journal writes. */ if (LIST_FIRST(&bmsafemap->sm_jaddrefhd) != NULL) { cgp = (struct cg *)bp->b_data; fs = VFSTOUFS(bmsafemap->sm_list.wk_mp)->um_fs; inosused = cg_inosused(cgp); LIST_FOREACH(jaddref, &bmsafemap->sm_jaddrefhd, ja_bmdeps) { ino = jaddref->ja_ino % fs->fs_ipg; if (isset(inosused, ino)) { if ((jaddref->ja_mode & IFMT) == IFDIR) cgp->cg_cs.cs_ndir--; cgp->cg_cs.cs_nifree++; clrbit(inosused, ino); jaddref->ja_state &= ~ATTACHED; jaddref->ja_state |= UNDONE; stat_jaddref++; } else panic("initiate_write_bmsafemap: inode %ju " "marked free", (uintmax_t)jaddref->ja_ino); } } /* * Clear any block allocations which are pending journal writes. */ if (LIST_FIRST(&bmsafemap->sm_jnewblkhd) != NULL) { cgp = (struct cg *)bp->b_data; fs = VFSTOUFS(bmsafemap->sm_list.wk_mp)->um_fs; blksfree = cg_blksfree(cgp); LIST_FOREACH(jnewblk, &bmsafemap->sm_jnewblkhd, jn_deps) { if (jnewblk_rollback(jnewblk, fs, cgp, blksfree)) continue; panic("initiate_write_bmsafemap: block %jd " "marked free", jnewblk->jn_blkno); } } /* * Move allocation lists to the written lists so they can be * cleared once the block write is complete. */ LIST_SWAP(&bmsafemap->sm_inodedephd, &bmsafemap->sm_inodedepwr, inodedep, id_deps); LIST_SWAP(&bmsafemap->sm_newblkhd, &bmsafemap->sm_newblkwr, newblk, nb_deps); LIST_SWAP(&bmsafemap->sm_freehd, &bmsafemap->sm_freewr, worklist, wk_list); } /* * This routine is called during the completion interrupt * service routine for a disk write (from the procedure called * by the device driver to inform the filesystem caches of * a request completion). It should be called early in this * procedure, before the block is made available to other * processes or other routines are called. * */ static void softdep_disk_write_complete(bp) struct buf *bp; /* describes the completed disk write */ { struct worklist *wk; struct worklist *owk; struct ufsmount *ump; struct workhead reattach; struct freeblks *freeblks; struct buf *sbp; /* * If an error occurred while doing the write, then the data * has not hit the disk and the dependencies cannot be unrolled. */ if ((bp->b_ioflags & BIO_ERROR) != 0 && (bp->b_flags & B_INVAL) == 0) return; if ((wk = LIST_FIRST(&bp->b_dep)) == NULL) return; ump = VFSTOUFS(wk->wk_mp); LIST_INIT(&reattach); /* * This lock must not be released anywhere in this code segment. */ sbp = NULL; owk = NULL; ACQUIRE_LOCK(ump); while ((wk = LIST_FIRST(&bp->b_dep)) != NULL) { WORKLIST_REMOVE(wk); atomic_add_long(&dep_write[wk->wk_type], 1); if (wk == owk) panic("duplicate worklist: %p\n", wk); owk = wk; switch (wk->wk_type) { case D_PAGEDEP: if (handle_written_filepage(WK_PAGEDEP(wk), bp)) WORKLIST_INSERT(&reattach, wk); continue; case D_INODEDEP: if (handle_written_inodeblock(WK_INODEDEP(wk), bp)) WORKLIST_INSERT(&reattach, wk); continue; case D_BMSAFEMAP: if (handle_written_bmsafemap(WK_BMSAFEMAP(wk), bp)) WORKLIST_INSERT(&reattach, wk); continue; case D_MKDIR: handle_written_mkdir(WK_MKDIR(wk), MKDIR_BODY); continue; case D_ALLOCDIRECT: wk->wk_state |= COMPLETE; handle_allocdirect_partdone(WK_ALLOCDIRECT(wk), NULL); continue; case D_ALLOCINDIR: wk->wk_state |= COMPLETE; handle_allocindir_partdone(WK_ALLOCINDIR(wk)); continue; case D_INDIRDEP: if (handle_written_indirdep(WK_INDIRDEP(wk), bp, &sbp)) WORKLIST_INSERT(&reattach, wk); continue; case D_FREEBLKS: wk->wk_state |= COMPLETE; freeblks = WK_FREEBLKS(wk); if ((wk->wk_state & ALLCOMPLETE) == ALLCOMPLETE && LIST_EMPTY(&freeblks->fb_jblkdephd)) add_to_worklist(wk, WK_NODELAY); continue; case D_FREEWORK: handle_written_freework(WK_FREEWORK(wk)); break; case D_JSEGDEP: free_jsegdep(WK_JSEGDEP(wk)); continue; case D_JSEG: handle_written_jseg(WK_JSEG(wk), bp); continue; case D_SBDEP: if (handle_written_sbdep(WK_SBDEP(wk), bp)) WORKLIST_INSERT(&reattach, wk); continue; case D_FREEDEP: free_freedep(WK_FREEDEP(wk)); continue; default: panic("handle_disk_write_complete: Unknown type %s", TYPENAME(wk->wk_type)); /* NOTREACHED */ } } /* * Reattach any requests that must be redone. */ while ((wk = LIST_FIRST(&reattach)) != NULL) { WORKLIST_REMOVE(wk); WORKLIST_INSERT(&bp->b_dep, wk); } FREE_LOCK(ump); if (sbp) brelse(sbp); } /* * Called from within softdep_disk_write_complete above. Note that * this routine is always called from interrupt level with further * splbio interrupts blocked. */ static void handle_allocdirect_partdone(adp, wkhd) struct allocdirect *adp; /* the completed allocdirect */ struct workhead *wkhd; /* Work to do when inode is writtne. */ { struct allocdirectlst *listhead; struct allocdirect *listadp; struct inodedep *inodedep; long bsize; if ((adp->ad_state & ALLCOMPLETE) != ALLCOMPLETE) return; /* * The on-disk inode cannot claim to be any larger than the last * fragment that has been written. Otherwise, the on-disk inode * might have fragments that were not the last block in the file * which would corrupt the filesystem. Thus, we cannot free any * allocdirects after one whose ad_oldblkno claims a fragment as * these blocks must be rolled back to zero before writing the inode. * We check the currently active set of allocdirects in id_inoupdt * or id_extupdt as appropriate. */ inodedep = adp->ad_inodedep; bsize = inodedep->id_fs->fs_bsize; if (adp->ad_state & EXTDATA) listhead = &inodedep->id_extupdt; else listhead = &inodedep->id_inoupdt; TAILQ_FOREACH(listadp, listhead, ad_next) { /* found our block */ if (listadp == adp) break; /* continue if ad_oldlbn is not a fragment */ if (listadp->ad_oldsize == 0 || listadp->ad_oldsize == bsize) continue; /* hit a fragment */ return; } /* * If we have reached the end of the current list without * finding the just finished dependency, then it must be * on the future dependency list. Future dependencies cannot * be freed until they are moved to the current list. */ if (listadp == NULL) { #ifdef DEBUG if (adp->ad_state & EXTDATA) listhead = &inodedep->id_newextupdt; else listhead = &inodedep->id_newinoupdt; TAILQ_FOREACH(listadp, listhead, ad_next) /* found our block */ if (listadp == adp) break; if (listadp == NULL) panic("handle_allocdirect_partdone: lost dep"); #endif /* DEBUG */ return; } /* * If we have found the just finished dependency, then queue * it along with anything that follows it that is complete. * Since the pointer has not yet been written in the inode * as the dependency prevents it, place the allocdirect on the * bufwait list where it will be freed once the pointer is * valid. */ if (wkhd == NULL) wkhd = &inodedep->id_bufwait; for (; adp; adp = listadp) { listadp = TAILQ_NEXT(adp, ad_next); if ((adp->ad_state & ALLCOMPLETE) != ALLCOMPLETE) return; TAILQ_REMOVE(listhead, adp, ad_next); WORKLIST_INSERT(wkhd, &adp->ad_block.nb_list); } } /* * Called from within softdep_disk_write_complete above. This routine * completes successfully written allocindirs. */ static void handle_allocindir_partdone(aip) struct allocindir *aip; /* the completed allocindir */ { struct indirdep *indirdep; if ((aip->ai_state & ALLCOMPLETE) != ALLCOMPLETE) return; indirdep = aip->ai_indirdep; LIST_REMOVE(aip, ai_next); /* * Don't set a pointer while the buffer is undergoing IO or while * we have active truncations. */ if (indirdep->ir_state & UNDONE || !TAILQ_EMPTY(&indirdep->ir_trunc)) { LIST_INSERT_HEAD(&indirdep->ir_donehd, aip, ai_next); return; } if (indirdep->ir_state & UFS1FMT) ((ufs1_daddr_t *)indirdep->ir_savebp->b_data)[aip->ai_offset] = aip->ai_newblkno; else ((ufs2_daddr_t *)indirdep->ir_savebp->b_data)[aip->ai_offset] = aip->ai_newblkno; /* * Await the pointer write before freeing the allocindir. */ LIST_INSERT_HEAD(&indirdep->ir_writehd, aip, ai_next); } /* * Release segments held on a jwork list. */ static void handle_jwork(wkhd) struct workhead *wkhd; { struct worklist *wk; while ((wk = LIST_FIRST(wkhd)) != NULL) { WORKLIST_REMOVE(wk); switch (wk->wk_type) { case D_JSEGDEP: free_jsegdep(WK_JSEGDEP(wk)); continue; case D_FREEDEP: free_freedep(WK_FREEDEP(wk)); continue; case D_FREEFRAG: rele_jseg(WK_JSEG(WK_FREEFRAG(wk)->ff_jdep)); WORKITEM_FREE(wk, D_FREEFRAG); continue; case D_FREEWORK: handle_written_freework(WK_FREEWORK(wk)); continue; default: panic("handle_jwork: Unknown type %s\n", TYPENAME(wk->wk_type)); } } } /* * Handle the bufwait list on an inode when it is safe to release items * held there. This normally happens after an inode block is written but * may be delayed and handled later if there are pending journal items that * are not yet safe to be released. */ static struct freefile * handle_bufwait(inodedep, refhd) struct inodedep *inodedep; struct workhead *refhd; { struct jaddref *jaddref; struct freefile *freefile; struct worklist *wk; freefile = NULL; while ((wk = LIST_FIRST(&inodedep->id_bufwait)) != NULL) { WORKLIST_REMOVE(wk); switch (wk->wk_type) { case D_FREEFILE: /* * We defer adding freefile to the worklist * until all other additions have been made to * ensure that it will be done after all the * old blocks have been freed. */ if (freefile != NULL) panic("handle_bufwait: freefile"); freefile = WK_FREEFILE(wk); continue; case D_MKDIR: handle_written_mkdir(WK_MKDIR(wk), MKDIR_PARENT); continue; case D_DIRADD: diradd_inode_written(WK_DIRADD(wk), inodedep); continue; case D_FREEFRAG: wk->wk_state |= COMPLETE; if ((wk->wk_state & ALLCOMPLETE) == ALLCOMPLETE) add_to_worklist(wk, 0); continue; case D_DIRREM: wk->wk_state |= COMPLETE; add_to_worklist(wk, 0); continue; case D_ALLOCDIRECT: case D_ALLOCINDIR: free_newblk(WK_NEWBLK(wk)); continue; case D_JNEWBLK: wk->wk_state |= COMPLETE; free_jnewblk(WK_JNEWBLK(wk)); continue; /* * Save freed journal segments and add references on * the supplied list which will delay their release * until the cg bitmap is cleared on disk. */ case D_JSEGDEP: if (refhd == NULL) free_jsegdep(WK_JSEGDEP(wk)); else WORKLIST_INSERT(refhd, wk); continue; case D_JADDREF: jaddref = WK_JADDREF(wk); TAILQ_REMOVE(&inodedep->id_inoreflst, &jaddref->ja_ref, if_deps); /* * Transfer any jaddrefs to the list to be freed with * the bitmap if we're handling a removed file. */ if (refhd == NULL) { wk->wk_state |= COMPLETE; free_jaddref(jaddref); } else WORKLIST_INSERT(refhd, wk); continue; default: panic("handle_bufwait: Unknown type %p(%s)", wk, TYPENAME(wk->wk_type)); /* NOTREACHED */ } } return (freefile); } /* * Called from within softdep_disk_write_complete above to restore * in-memory inode block contents to their most up-to-date state. Note * that this routine is always called from interrupt level with further * splbio interrupts blocked. */ static int handle_written_inodeblock(inodedep, bp) struct inodedep *inodedep; struct buf *bp; /* buffer containing the inode block */ { struct freefile *freefile; struct allocdirect *adp, *nextadp; struct ufs1_dinode *dp1 = NULL; struct ufs2_dinode *dp2 = NULL; struct workhead wkhd; int hadchanges, fstype; ino_t freelink; LIST_INIT(&wkhd); hadchanges = 0; freefile = NULL; if ((inodedep->id_state & IOSTARTED) == 0) panic("handle_written_inodeblock: not started"); inodedep->id_state &= ~IOSTARTED; if (inodedep->id_fs->fs_magic == FS_UFS1_MAGIC) { fstype = UFS1; dp1 = (struct ufs1_dinode *)bp->b_data + ino_to_fsbo(inodedep->id_fs, inodedep->id_ino); freelink = dp1->di_freelink; } else { fstype = UFS2; dp2 = (struct ufs2_dinode *)bp->b_data + ino_to_fsbo(inodedep->id_fs, inodedep->id_ino); freelink = dp2->di_freelink; } /* * Leave this inodeblock dirty until it's in the list. */ if ((inodedep->id_state & (UNLINKED | UNLINKONLIST)) == UNLINKED) { struct inodedep *inon; inon = TAILQ_NEXT(inodedep, id_unlinked); if ((inon == NULL && freelink == 0) || (inon && inon->id_ino == freelink)) { if (inon) inon->id_state |= UNLINKPREV; inodedep->id_state |= UNLINKNEXT; } hadchanges = 1; } /* * If we had to rollback the inode allocation because of * bitmaps being incomplete, then simply restore it. * Keep the block dirty so that it will not be reclaimed until * all associated dependencies have been cleared and the * corresponding updates written to disk. */ if (inodedep->id_savedino1 != NULL) { hadchanges = 1; if (fstype == UFS1) *dp1 = *inodedep->id_savedino1; else *dp2 = *inodedep->id_savedino2; free(inodedep->id_savedino1, M_SAVEDINO); inodedep->id_savedino1 = NULL; if ((bp->b_flags & B_DELWRI) == 0) stat_inode_bitmap++; bdirty(bp); /* * If the inode is clear here and GOINGAWAY it will never * be written. Process the bufwait and clear any pending * work which may include the freefile. */ if (inodedep->id_state & GOINGAWAY) goto bufwait; return (1); } inodedep->id_state |= COMPLETE; /* * Roll forward anything that had to be rolled back before * the inode could be updated. */ for (adp = TAILQ_FIRST(&inodedep->id_inoupdt); adp; adp = nextadp) { nextadp = TAILQ_NEXT(adp, ad_next); if (adp->ad_state & ATTACHED) panic("handle_written_inodeblock: new entry"); if (fstype == UFS1) { if (adp->ad_offset < NDADDR) { if (dp1->di_db[adp->ad_offset]!=adp->ad_oldblkno) panic("%s %s #%jd mismatch %d != %jd", "handle_written_inodeblock:", "direct pointer", (intmax_t)adp->ad_offset, dp1->di_db[adp->ad_offset], (intmax_t)adp->ad_oldblkno); dp1->di_db[adp->ad_offset] = adp->ad_newblkno; } else { if (dp1->di_ib[adp->ad_offset - NDADDR] != 0) panic("%s: %s #%jd allocated as %d", "handle_written_inodeblock", "indirect pointer", (intmax_t)adp->ad_offset - NDADDR, dp1->di_ib[adp->ad_offset - NDADDR]); dp1->di_ib[adp->ad_offset - NDADDR] = adp->ad_newblkno; } } else { if (adp->ad_offset < NDADDR) { if (dp2->di_db[adp->ad_offset]!=adp->ad_oldblkno) panic("%s: %s #%jd %s %jd != %jd", "handle_written_inodeblock", "direct pointer", (intmax_t)adp->ad_offset, "mismatch", (intmax_t)dp2->di_db[adp->ad_offset], (intmax_t)adp->ad_oldblkno); dp2->di_db[adp->ad_offset] = adp->ad_newblkno; } else { if (dp2->di_ib[adp->ad_offset - NDADDR] != 0) panic("%s: %s #%jd allocated as %jd", "handle_written_inodeblock", "indirect pointer", (intmax_t)adp->ad_offset - NDADDR, (intmax_t) dp2->di_ib[adp->ad_offset - NDADDR]); dp2->di_ib[adp->ad_offset - NDADDR] = adp->ad_newblkno; } } adp->ad_state &= ~UNDONE; adp->ad_state |= ATTACHED; hadchanges = 1; } for (adp = TAILQ_FIRST(&inodedep->id_extupdt); adp; adp = nextadp) { nextadp = TAILQ_NEXT(adp, ad_next); if (adp->ad_state & ATTACHED) panic("handle_written_inodeblock: new entry"); if (dp2->di_extb[adp->ad_offset] != adp->ad_oldblkno) panic("%s: direct pointers #%jd %s %jd != %jd", "handle_written_inodeblock", (intmax_t)adp->ad_offset, "mismatch", (intmax_t)dp2->di_extb[adp->ad_offset], (intmax_t)adp->ad_oldblkno); dp2->di_extb[adp->ad_offset] = adp->ad_newblkno; adp->ad_state &= ~UNDONE; adp->ad_state |= ATTACHED; hadchanges = 1; } if (hadchanges && (bp->b_flags & B_DELWRI) == 0) stat_direct_blk_ptrs++; /* * Reset the file size to its most up-to-date value. */ if (inodedep->id_savedsize == -1 || inodedep->id_savedextsize == -1) panic("handle_written_inodeblock: bad size"); if (inodedep->id_savednlink > LINK_MAX) panic("handle_written_inodeblock: Invalid link count " "%d for inodedep %p", inodedep->id_savednlink, inodedep); if (fstype == UFS1) { if (dp1->di_nlink != inodedep->id_savednlink) { dp1->di_nlink = inodedep->id_savednlink; hadchanges = 1; } if (dp1->di_size != inodedep->id_savedsize) { dp1->di_size = inodedep->id_savedsize; hadchanges = 1; } } else { if (dp2->di_nlink != inodedep->id_savednlink) { dp2->di_nlink = inodedep->id_savednlink; hadchanges = 1; } if (dp2->di_size != inodedep->id_savedsize) { dp2->di_size = inodedep->id_savedsize; hadchanges = 1; } if (dp2->di_extsize != inodedep->id_savedextsize) { dp2->di_extsize = inodedep->id_savedextsize; hadchanges = 1; } } inodedep->id_savedsize = -1; inodedep->id_savedextsize = -1; inodedep->id_savednlink = -1; /* * If there were any rollbacks in the inode block, then it must be * marked dirty so that its will eventually get written back in * its correct form. */ if (hadchanges) bdirty(bp); bufwait: /* * Process any allocdirects that completed during the update. */ if ((adp = TAILQ_FIRST(&inodedep->id_inoupdt)) != NULL) handle_allocdirect_partdone(adp, &wkhd); if ((adp = TAILQ_FIRST(&inodedep->id_extupdt)) != NULL) handle_allocdirect_partdone(adp, &wkhd); /* * Process deallocations that were held pending until the * inode had been written to disk. Freeing of the inode * is delayed until after all blocks have been freed to * avoid creation of new triples * before the old ones have been deleted. Completely * unlinked inodes are not processed until the unlinked * inode list is written or the last reference is removed. */ if ((inodedep->id_state & (UNLINKED | UNLINKONLIST)) != UNLINKED) { freefile = handle_bufwait(inodedep, NULL); if (freefile && !LIST_EMPTY(&wkhd)) { WORKLIST_INSERT(&wkhd, &freefile->fx_list); freefile = NULL; } } /* * Move rolled forward dependency completions to the bufwait list * now that those that were already written have been processed. */ if (!LIST_EMPTY(&wkhd) && hadchanges == 0) panic("handle_written_inodeblock: bufwait but no changes"); jwork_move(&inodedep->id_bufwait, &wkhd); if (freefile != NULL) { /* * If the inode is goingaway it was never written. Fake up * the state here so free_inodedep() can succeed. */ if (inodedep->id_state & GOINGAWAY) inodedep->id_state |= COMPLETE | DEPCOMPLETE; if (free_inodedep(inodedep) == 0) panic("handle_written_inodeblock: live inodedep %p", inodedep); add_to_worklist(&freefile->fx_list, 0); return (0); } /* * If no outstanding dependencies, free it. */ if (free_inodedep(inodedep) || (TAILQ_FIRST(&inodedep->id_inoreflst) == 0 && TAILQ_FIRST(&inodedep->id_inoupdt) == 0 && TAILQ_FIRST(&inodedep->id_extupdt) == 0 && LIST_FIRST(&inodedep->id_bufwait) == 0)) return (0); return (hadchanges); } static int handle_written_indirdep(indirdep, bp, bpp) struct indirdep *indirdep; struct buf *bp; struct buf **bpp; { struct allocindir *aip; struct buf *sbp; int chgs; if (indirdep->ir_state & GOINGAWAY) panic("handle_written_indirdep: indirdep gone"); if ((indirdep->ir_state & IOSTARTED) == 0) panic("handle_written_indirdep: IO not started"); chgs = 0; /* * If there were rollbacks revert them here. */ if (indirdep->ir_saveddata) { bcopy(indirdep->ir_saveddata, bp->b_data, bp->b_bcount); if (TAILQ_EMPTY(&indirdep->ir_trunc)) { free(indirdep->ir_saveddata, M_INDIRDEP); indirdep->ir_saveddata = NULL; } chgs = 1; } indirdep->ir_state &= ~(UNDONE | IOSTARTED); indirdep->ir_state |= ATTACHED; /* * Move allocindirs with written pointers to the completehd if * the indirdep's pointer is not yet written. Otherwise * free them here. */ while ((aip = LIST_FIRST(&indirdep->ir_writehd)) != 0) { LIST_REMOVE(aip, ai_next); if ((indirdep->ir_state & DEPCOMPLETE) == 0) { LIST_INSERT_HEAD(&indirdep->ir_completehd, aip, ai_next); newblk_freefrag(&aip->ai_block); continue; } free_newblk(&aip->ai_block); } /* * Move allocindirs that have finished dependency processing from * the done list to the write list after updating the pointers. */ if (TAILQ_EMPTY(&indirdep->ir_trunc)) { while ((aip = LIST_FIRST(&indirdep->ir_donehd)) != 0) { handle_allocindir_partdone(aip); if (aip == LIST_FIRST(&indirdep->ir_donehd)) panic("disk_write_complete: not gone"); chgs = 1; } } /* * Preserve the indirdep if there were any changes or if it is not * yet valid on disk. */ if (chgs) { stat_indir_blk_ptrs++; bdirty(bp); return (1); } /* * If there were no changes we can discard the savedbp and detach * ourselves from the buf. We are only carrying completed pointers * in this case. */ sbp = indirdep->ir_savebp; sbp->b_flags |= B_INVAL | B_NOCACHE; indirdep->ir_savebp = NULL; indirdep->ir_bp = NULL; if (*bpp != NULL) panic("handle_written_indirdep: bp already exists."); *bpp = sbp; /* * The indirdep may not be freed until its parent points at it. */ if (indirdep->ir_state & DEPCOMPLETE) free_indirdep(indirdep); return (0); } /* * Process a diradd entry after its dependent inode has been written. * This routine must be called with splbio interrupts blocked. */ static void diradd_inode_written(dap, inodedep) struct diradd *dap; struct inodedep *inodedep; { dap->da_state |= COMPLETE; complete_diradd(dap); WORKLIST_INSERT(&inodedep->id_pendinghd, &dap->da_list); } /* * Returns true if the bmsafemap will have rollbacks when written. Must only * be called with the per-filesystem lock and the buf lock on the cg held. */ static int bmsafemap_backgroundwrite(bmsafemap, bp) struct bmsafemap *bmsafemap; struct buf *bp; { int dirty; LOCK_OWNED(VFSTOUFS(bmsafemap->sm_list.wk_mp)); dirty = !LIST_EMPTY(&bmsafemap->sm_jaddrefhd) | !LIST_EMPTY(&bmsafemap->sm_jnewblkhd); /* * If we're initiating a background write we need to process the * rollbacks as they exist now, not as they exist when IO starts. * No other consumers will look at the contents of the shadowed * buf so this is safe to do here. */ if (bp->b_xflags & BX_BKGRDMARKER) initiate_write_bmsafemap(bmsafemap, bp); return (dirty); } /* * Re-apply an allocation when a cg write is complete. */ static int jnewblk_rollforward(jnewblk, fs, cgp, blksfree) struct jnewblk *jnewblk; struct fs *fs; struct cg *cgp; uint8_t *blksfree; { ufs1_daddr_t fragno; ufs2_daddr_t blkno; long cgbno, bbase; int frags, blk; int i; frags = 0; cgbno = dtogd(fs, jnewblk->jn_blkno); for (i = jnewblk->jn_oldfrags; i < jnewblk->jn_frags; i++) { if (isclr(blksfree, cgbno + i)) panic("jnewblk_rollforward: re-allocated fragment"); frags++; } if (frags == fs->fs_frag) { blkno = fragstoblks(fs, cgbno); ffs_clrblock(fs, blksfree, (long)blkno); ffs_clusteracct(fs, cgp, blkno, -1); cgp->cg_cs.cs_nbfree--; } else { bbase = cgbno - fragnum(fs, cgbno); cgbno += jnewblk->jn_oldfrags; /* If a complete block had been reassembled, account for it. */ fragno = fragstoblks(fs, bbase); if (ffs_isblock(fs, blksfree, fragno)) { cgp->cg_cs.cs_nffree += fs->fs_frag; ffs_clusteracct(fs, cgp, fragno, -1); cgp->cg_cs.cs_nbfree--; } /* Decrement the old frags. */ blk = blkmap(fs, blksfree, bbase); ffs_fragacct(fs, blk, cgp->cg_frsum, -1); /* Allocate the fragment */ for (i = 0; i < frags; i++) clrbit(blksfree, cgbno + i); cgp->cg_cs.cs_nffree -= frags; /* Add back in counts associated with the new frags */ blk = blkmap(fs, blksfree, bbase); ffs_fragacct(fs, blk, cgp->cg_frsum, 1); } return (frags); } /* * Complete a write to a bmsafemap structure. Roll forward any bitmap * changes if it's not a background write. Set all written dependencies * to DEPCOMPLETE and free the structure if possible. */ static int handle_written_bmsafemap(bmsafemap, bp) struct bmsafemap *bmsafemap; struct buf *bp; { struct newblk *newblk; struct inodedep *inodedep; struct jaddref *jaddref, *jatmp; struct jnewblk *jnewblk, *jntmp; struct ufsmount *ump; uint8_t *inosused; uint8_t *blksfree; struct cg *cgp; struct fs *fs; ino_t ino; int foreground; int chgs; if ((bmsafemap->sm_state & IOSTARTED) == 0) panic("initiate_write_bmsafemap: Not started\n"); ump = VFSTOUFS(bmsafemap->sm_list.wk_mp); chgs = 0; bmsafemap->sm_state &= ~IOSTARTED; foreground = (bp->b_xflags & BX_BKGRDMARKER) == 0; /* * Release journal work that was waiting on the write. */ handle_jwork(&bmsafemap->sm_freewr); /* * Restore unwritten inode allocation pending jaddref writes. */ if (!LIST_EMPTY(&bmsafemap->sm_jaddrefhd)) { cgp = (struct cg *)bp->b_data; fs = VFSTOUFS(bmsafemap->sm_list.wk_mp)->um_fs; inosused = cg_inosused(cgp); LIST_FOREACH_SAFE(jaddref, &bmsafemap->sm_jaddrefhd, ja_bmdeps, jatmp) { if ((jaddref->ja_state & UNDONE) == 0) continue; ino = jaddref->ja_ino % fs->fs_ipg; if (isset(inosused, ino)) panic("handle_written_bmsafemap: " "re-allocated inode"); /* Do the roll-forward only if it's a real copy. */ if (foreground) { if ((jaddref->ja_mode & IFMT) == IFDIR) cgp->cg_cs.cs_ndir++; cgp->cg_cs.cs_nifree--; setbit(inosused, ino); chgs = 1; } jaddref->ja_state &= ~UNDONE; jaddref->ja_state |= ATTACHED; free_jaddref(jaddref); } } /* * Restore any block allocations which are pending journal writes. */ if (LIST_FIRST(&bmsafemap->sm_jnewblkhd) != NULL) { cgp = (struct cg *)bp->b_data; fs = VFSTOUFS(bmsafemap->sm_list.wk_mp)->um_fs; blksfree = cg_blksfree(cgp); LIST_FOREACH_SAFE(jnewblk, &bmsafemap->sm_jnewblkhd, jn_deps, jntmp) { if ((jnewblk->jn_state & UNDONE) == 0) continue; /* Do the roll-forward only if it's a real copy. */ if (foreground && jnewblk_rollforward(jnewblk, fs, cgp, blksfree)) chgs = 1; jnewblk->jn_state &= ~(UNDONE | NEWBLOCK); jnewblk->jn_state |= ATTACHED; free_jnewblk(jnewblk); } } while ((newblk = LIST_FIRST(&bmsafemap->sm_newblkwr))) { newblk->nb_state |= DEPCOMPLETE; newblk->nb_state &= ~ONDEPLIST; newblk->nb_bmsafemap = NULL; LIST_REMOVE(newblk, nb_deps); if (newblk->nb_list.wk_type == D_ALLOCDIRECT) handle_allocdirect_partdone( WK_ALLOCDIRECT(&newblk->nb_list), NULL); else if (newblk->nb_list.wk_type == D_ALLOCINDIR) handle_allocindir_partdone( WK_ALLOCINDIR(&newblk->nb_list)); else if (newblk->nb_list.wk_type != D_NEWBLK) panic("handle_written_bmsafemap: Unexpected type: %s", TYPENAME(newblk->nb_list.wk_type)); } while ((inodedep = LIST_FIRST(&bmsafemap->sm_inodedepwr)) != NULL) { inodedep->id_state |= DEPCOMPLETE; inodedep->id_state &= ~ONDEPLIST; LIST_REMOVE(inodedep, id_deps); inodedep->id_bmsafemap = NULL; } LIST_REMOVE(bmsafemap, sm_next); if (chgs == 0 && LIST_EMPTY(&bmsafemap->sm_jaddrefhd) && LIST_EMPTY(&bmsafemap->sm_jnewblkhd) && LIST_EMPTY(&bmsafemap->sm_newblkhd) && LIST_EMPTY(&bmsafemap->sm_inodedephd) && LIST_EMPTY(&bmsafemap->sm_freehd)) { LIST_REMOVE(bmsafemap, sm_hash); WORKITEM_FREE(bmsafemap, D_BMSAFEMAP); return (0); } LIST_INSERT_HEAD(&ump->softdep_dirtycg, bmsafemap, sm_next); if (foreground) bdirty(bp); return (1); } /* * Try to free a mkdir dependency. */ static void complete_mkdir(mkdir) struct mkdir *mkdir; { struct diradd *dap; if ((mkdir->md_state & ALLCOMPLETE) != ALLCOMPLETE) return; LIST_REMOVE(mkdir, md_mkdirs); dap = mkdir->md_diradd; dap->da_state &= ~(mkdir->md_state & (MKDIR_PARENT | MKDIR_BODY)); if ((dap->da_state & (MKDIR_PARENT | MKDIR_BODY)) == 0) { dap->da_state |= DEPCOMPLETE; complete_diradd(dap); } WORKITEM_FREE(mkdir, D_MKDIR); } /* * Handle the completion of a mkdir dependency. */ static void handle_written_mkdir(mkdir, type) struct mkdir *mkdir; int type; { if ((mkdir->md_state & (MKDIR_PARENT | MKDIR_BODY)) != type) panic("handle_written_mkdir: bad type"); mkdir->md_state |= COMPLETE; complete_mkdir(mkdir); } static int free_pagedep(pagedep) struct pagedep *pagedep; { int i; if (pagedep->pd_state & NEWBLOCK) return (0); if (!LIST_EMPTY(&pagedep->pd_dirremhd)) return (0); for (i = 0; i < DAHASHSZ; i++) if (!LIST_EMPTY(&pagedep->pd_diraddhd[i])) return (0); if (!LIST_EMPTY(&pagedep->pd_pendinghd)) return (0); if (!LIST_EMPTY(&pagedep->pd_jmvrefhd)) return (0); if (pagedep->pd_state & ONWORKLIST) WORKLIST_REMOVE(&pagedep->pd_list); LIST_REMOVE(pagedep, pd_hash); WORKITEM_FREE(pagedep, D_PAGEDEP); return (1); } /* * Called from within softdep_disk_write_complete above. * A write operation was just completed. Removed inodes can * now be freed and associated block pointers may be committed. * Note that this routine is always called from interrupt level * with further splbio interrupts blocked. */ static int handle_written_filepage(pagedep, bp) struct pagedep *pagedep; struct buf *bp; /* buffer containing the written page */ { struct dirrem *dirrem; struct diradd *dap, *nextdap; struct direct *ep; int i, chgs; if ((pagedep->pd_state & IOSTARTED) == 0) panic("handle_written_filepage: not started"); pagedep->pd_state &= ~IOSTARTED; /* * Process any directory removals that have been committed. */ while ((dirrem = LIST_FIRST(&pagedep->pd_dirremhd)) != NULL) { LIST_REMOVE(dirrem, dm_next); dirrem->dm_state |= COMPLETE; dirrem->dm_dirinum = pagedep->pd_ino; KASSERT(LIST_EMPTY(&dirrem->dm_jremrefhd), ("handle_written_filepage: Journal entries not written.")); add_to_worklist(&dirrem->dm_list, 0); } /* * Free any directory additions that have been committed. * If it is a newly allocated block, we have to wait until * the on-disk directory inode claims the new block. */ if ((pagedep->pd_state & NEWBLOCK) == 0) while ((dap = LIST_FIRST(&pagedep->pd_pendinghd)) != NULL) free_diradd(dap, NULL); /* * Uncommitted directory entries must be restored. */ for (chgs = 0, i = 0; i < DAHASHSZ; i++) { for (dap = LIST_FIRST(&pagedep->pd_diraddhd[i]); dap; dap = nextdap) { nextdap = LIST_NEXT(dap, da_pdlist); if (dap->da_state & ATTACHED) panic("handle_written_filepage: attached"); ep = (struct direct *) ((char *)bp->b_data + dap->da_offset); ep->d_ino = dap->da_newinum; dap->da_state &= ~UNDONE; dap->da_state |= ATTACHED; chgs = 1; /* * If the inode referenced by the directory has * been written out, then the dependency can be * moved to the pending list. */ if ((dap->da_state & ALLCOMPLETE) == ALLCOMPLETE) { LIST_REMOVE(dap, da_pdlist); LIST_INSERT_HEAD(&pagedep->pd_pendinghd, dap, da_pdlist); } } } /* * If there were any rollbacks in the directory, then it must be * marked dirty so that its will eventually get written back in * its correct form. */ if (chgs) { if ((bp->b_flags & B_DELWRI) == 0) stat_dir_entry++; bdirty(bp); return (1); } /* * If we are not waiting for a new directory block to be * claimed by its inode, then the pagedep will be freed. * Otherwise it will remain to track any new entries on * the page in case they are fsync'ed. */ free_pagedep(pagedep); return (0); } /* * Writing back in-core inode structures. * * The filesystem only accesses an inode's contents when it occupies an * "in-core" inode structure. These "in-core" structures are separate from * the page frames used to cache inode blocks. Only the latter are * transferred to/from the disk. So, when the updated contents of the * "in-core" inode structure are copied to the corresponding in-memory inode * block, the dependencies are also transferred. The following procedure is * called when copying a dirty "in-core" inode to a cached inode block. */ /* * Called when an inode is loaded from disk. If the effective link count * differed from the actual link count when it was last flushed, then we * need to ensure that the correct effective link count is put back. */ void softdep_load_inodeblock(ip) struct inode *ip; /* the "in_core" copy of the inode */ { struct inodedep *inodedep; KASSERT(MOUNTEDSOFTDEP(UFSTOVFS(ip->i_ump)) != 0, ("softdep_load_inodeblock called on non-softdep filesystem")); /* * Check for alternate nlink count. */ ip->i_effnlink = ip->i_nlink; ACQUIRE_LOCK(ip->i_ump); if (inodedep_lookup(UFSTOVFS(ip->i_ump), ip->i_number, 0, &inodedep) == 0) { FREE_LOCK(ip->i_ump); return; } ip->i_effnlink -= inodedep->id_nlinkdelta; FREE_LOCK(ip->i_ump); } /* * This routine is called just before the "in-core" inode * information is to be copied to the in-memory inode block. * Recall that an inode block contains several inodes. If * the force flag is set, then the dependencies will be * cleared so that the update can always be made. Note that * the buffer is locked when this routine is called, so we * will never be in the middle of writing the inode block * to disk. */ void softdep_update_inodeblock(ip, bp, waitfor) struct inode *ip; /* the "in_core" copy of the inode */ struct buf *bp; /* the buffer containing the inode block */ int waitfor; /* nonzero => update must be allowed */ { struct inodedep *inodedep; struct inoref *inoref; struct ufsmount *ump; struct worklist *wk; struct mount *mp; struct buf *ibp; struct fs *fs; int error; ump = ip->i_ump; mp = UFSTOVFS(ump); KASSERT(MOUNTEDSOFTDEP(mp) != 0, ("softdep_update_inodeblock called on non-softdep filesystem")); fs = ip->i_fs; /* * Preserve the freelink that is on disk. clear_unlinked_inodedep() * does not have access to the in-core ip so must write directly into * the inode block buffer when setting freelink. */ if (fs->fs_magic == FS_UFS1_MAGIC) DIP_SET(ip, i_freelink, ((struct ufs1_dinode *)bp->b_data + ino_to_fsbo(fs, ip->i_number))->di_freelink); else DIP_SET(ip, i_freelink, ((struct ufs2_dinode *)bp->b_data + ino_to_fsbo(fs, ip->i_number))->di_freelink); /* * If the effective link count is not equal to the actual link * count, then we must track the difference in an inodedep while * the inode is (potentially) tossed out of the cache. Otherwise, * if there is no existing inodedep, then there are no dependencies * to track. */ ACQUIRE_LOCK(ump); again: if (inodedep_lookup(mp, ip->i_number, 0, &inodedep) == 0) { FREE_LOCK(ump); if (ip->i_effnlink != ip->i_nlink) panic("softdep_update_inodeblock: bad link count"); return; } if (inodedep->id_nlinkdelta != ip->i_nlink - ip->i_effnlink) panic("softdep_update_inodeblock: bad delta"); /* * If we're flushing all dependencies we must also move any waiting * for journal writes onto the bufwait list prior to I/O. */ if (waitfor) { TAILQ_FOREACH(inoref, &inodedep->id_inoreflst, if_deps) { if ((inoref->if_state & (DEPCOMPLETE | GOINGAWAY)) == DEPCOMPLETE) { jwait(&inoref->if_list, MNT_WAIT); goto again; } } } /* * Changes have been initiated. Anything depending on these * changes cannot occur until this inode has been written. */ inodedep->id_state &= ~COMPLETE; if ((inodedep->id_state & ONWORKLIST) == 0) WORKLIST_INSERT(&bp->b_dep, &inodedep->id_list); /* * Any new dependencies associated with the incore inode must * now be moved to the list associated with the buffer holding * the in-memory copy of the inode. Once merged process any * allocdirects that are completed by the merger. */ merge_inode_lists(&inodedep->id_newinoupdt, &inodedep->id_inoupdt); if (!TAILQ_EMPTY(&inodedep->id_inoupdt)) handle_allocdirect_partdone(TAILQ_FIRST(&inodedep->id_inoupdt), NULL); merge_inode_lists(&inodedep->id_newextupdt, &inodedep->id_extupdt); if (!TAILQ_EMPTY(&inodedep->id_extupdt)) handle_allocdirect_partdone(TAILQ_FIRST(&inodedep->id_extupdt), NULL); /* * Now that the inode has been pushed into the buffer, the * operations dependent on the inode being written to disk * can be moved to the id_bufwait so that they will be * processed when the buffer I/O completes. */ while ((wk = LIST_FIRST(&inodedep->id_inowait)) != NULL) { WORKLIST_REMOVE(wk); WORKLIST_INSERT(&inodedep->id_bufwait, wk); } /* * Newly allocated inodes cannot be written until the bitmap * that allocates them have been written (indicated by * DEPCOMPLETE being set in id_state). If we are doing a * forced sync (e.g., an fsync on a file), we force the bitmap * to be written so that the update can be done. */ if (waitfor == 0) { FREE_LOCK(ump); return; } retry: if ((inodedep->id_state & (DEPCOMPLETE | GOINGAWAY)) != 0) { FREE_LOCK(ump); return; } ibp = inodedep->id_bmsafemap->sm_buf; ibp = getdirtybuf(ibp, LOCK_PTR(ump), MNT_WAIT); if (ibp == NULL) { /* * If ibp came back as NULL, the dependency could have been * freed while we slept. Look it up again, and check to see * that it has completed. */ if (inodedep_lookup(mp, ip->i_number, 0, &inodedep) != 0) goto retry; FREE_LOCK(ump); return; } FREE_LOCK(ump); if ((error = bwrite(ibp)) != 0) softdep_error("softdep_update_inodeblock: bwrite", error); } /* * Merge the a new inode dependency list (such as id_newinoupdt) into an * old inode dependency list (such as id_inoupdt). This routine must be * called with splbio interrupts blocked. */ static void merge_inode_lists(newlisthead, oldlisthead) struct allocdirectlst *newlisthead; struct allocdirectlst *oldlisthead; { struct allocdirect *listadp, *newadp; newadp = TAILQ_FIRST(newlisthead); for (listadp = TAILQ_FIRST(oldlisthead); listadp && newadp;) { if (listadp->ad_offset < newadp->ad_offset) { listadp = TAILQ_NEXT(listadp, ad_next); continue; } TAILQ_REMOVE(newlisthead, newadp, ad_next); TAILQ_INSERT_BEFORE(listadp, newadp, ad_next); if (listadp->ad_offset == newadp->ad_offset) { allocdirect_merge(oldlisthead, newadp, listadp); listadp = newadp; } newadp = TAILQ_FIRST(newlisthead); } while ((newadp = TAILQ_FIRST(newlisthead)) != NULL) { TAILQ_REMOVE(newlisthead, newadp, ad_next); TAILQ_INSERT_TAIL(oldlisthead, newadp, ad_next); } } /* * If we are doing an fsync, then we must ensure that any directory * entries for the inode have been written after the inode gets to disk. */ int softdep_fsync(vp) struct vnode *vp; /* the "in_core" copy of the inode */ { struct inodedep *inodedep; struct pagedep *pagedep; struct inoref *inoref; struct ufsmount *ump; struct worklist *wk; struct diradd *dap; struct mount *mp; struct vnode *pvp; struct inode *ip; struct buf *bp; struct fs *fs; struct thread *td = curthread; int error, flushparent, pagedep_new_block; ino_t parentino; ufs_lbn_t lbn; ip = VTOI(vp); fs = ip->i_fs; ump = ip->i_ump; mp = vp->v_mount; if (MOUNTEDSOFTDEP(mp) == 0) return (0); ACQUIRE_LOCK(ump); restart: if (inodedep_lookup(mp, ip->i_number, 0, &inodedep) == 0) { FREE_LOCK(ump); return (0); } TAILQ_FOREACH(inoref, &inodedep->id_inoreflst, if_deps) { if ((inoref->if_state & (DEPCOMPLETE | GOINGAWAY)) == DEPCOMPLETE) { jwait(&inoref->if_list, MNT_WAIT); goto restart; } } if (!LIST_EMPTY(&inodedep->id_inowait) || !TAILQ_EMPTY(&inodedep->id_extupdt) || !TAILQ_EMPTY(&inodedep->id_newextupdt) || !TAILQ_EMPTY(&inodedep->id_inoupdt) || !TAILQ_EMPTY(&inodedep->id_newinoupdt)) panic("softdep_fsync: pending ops %p", inodedep); for (error = 0, flushparent = 0; ; ) { if ((wk = LIST_FIRST(&inodedep->id_pendinghd)) == NULL) break; if (wk->wk_type != D_DIRADD) panic("softdep_fsync: Unexpected type %s", TYPENAME(wk->wk_type)); dap = WK_DIRADD(wk); /* * Flush our parent if this directory entry has a MKDIR_PARENT * dependency or is contained in a newly allocated block. */ if (dap->da_state & DIRCHG) pagedep = dap->da_previous->dm_pagedep; else pagedep = dap->da_pagedep; parentino = pagedep->pd_ino; lbn = pagedep->pd_lbn; if ((dap->da_state & (MKDIR_BODY | COMPLETE)) != COMPLETE) panic("softdep_fsync: dirty"); if ((dap->da_state & MKDIR_PARENT) || (pagedep->pd_state & NEWBLOCK)) flushparent = 1; else flushparent = 0; /* * If we are being fsync'ed as part of vgone'ing this vnode, * then we will not be able to release and recover the * vnode below, so we just have to give up on writing its * directory entry out. It will eventually be written, just * not now, but then the user was not asking to have it * written, so we are not breaking any promises. */ if (vp->v_iflag & VI_DOOMED) break; /* * We prevent deadlock by always fetching inodes from the * root, moving down the directory tree. Thus, when fetching * our parent directory, we first try to get the lock. If * that fails, we must unlock ourselves before requesting * the lock on our parent. See the comment in ufs_lookup * for details on possible races. */ FREE_LOCK(ump); if (ffs_vgetf(mp, parentino, LK_NOWAIT | LK_EXCLUSIVE, &pvp, FFSV_FORCEINSMQ)) { error = vfs_busy(mp, MBF_NOWAIT); if (error != 0) { vfs_ref(mp); VOP_UNLOCK(vp, 0); error = vfs_busy(mp, 0); vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); vfs_rel(mp); if (error != 0) return (ENOENT); if (vp->v_iflag & VI_DOOMED) { vfs_unbusy(mp); return (ENOENT); } } VOP_UNLOCK(vp, 0); error = ffs_vgetf(mp, parentino, LK_EXCLUSIVE, &pvp, FFSV_FORCEINSMQ); vfs_unbusy(mp); vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); if (vp->v_iflag & VI_DOOMED) { if (error == 0) vput(pvp); error = ENOENT; } if (error != 0) return (error); } /* * All MKDIR_PARENT dependencies and all the NEWBLOCK pagedeps * that are contained in direct blocks will be resolved by * doing a ffs_update. Pagedeps contained in indirect blocks * may require a complete sync'ing of the directory. So, we * try the cheap and fast ffs_update first, and if that fails, * then we do the slower ffs_syncvnode of the directory. */ if (flushparent) { int locked; if ((error = ffs_update(pvp, 1)) != 0) { vput(pvp); return (error); } ACQUIRE_LOCK(ump); locked = 1; if (inodedep_lookup(mp, ip->i_number, 0, &inodedep) != 0) { if ((wk = LIST_FIRST(&inodedep->id_pendinghd)) != NULL) { if (wk->wk_type != D_DIRADD) panic("softdep_fsync: Unexpected type %s", TYPENAME(wk->wk_type)); dap = WK_DIRADD(wk); if (dap->da_state & DIRCHG) pagedep = dap->da_previous->dm_pagedep; else pagedep = dap->da_pagedep; pagedep_new_block = pagedep->pd_state & NEWBLOCK; FREE_LOCK(ump); locked = 0; if (pagedep_new_block && (error = ffs_syncvnode(pvp, MNT_WAIT, 0))) { vput(pvp); return (error); } } } if (locked) FREE_LOCK(ump); } /* * Flush directory page containing the inode's name. */ error = bread(pvp, lbn, blksize(fs, VTOI(pvp), lbn), td->td_ucred, &bp); if (error == 0) error = bwrite(bp); else brelse(bp); vput(pvp); if (error != 0) return (error); ACQUIRE_LOCK(ump); if (inodedep_lookup(mp, ip->i_number, 0, &inodedep) == 0) break; } FREE_LOCK(ump); return (0); } /* * Flush all the dirty bitmaps associated with the block device * before flushing the rest of the dirty blocks so as to reduce * the number of dependencies that will have to be rolled back. * * XXX Unused? */ void softdep_fsync_mountdev(vp) struct vnode *vp; { struct buf *bp, *nbp; struct worklist *wk; struct bufobj *bo; if (!vn_isdisk(vp, NULL)) panic("softdep_fsync_mountdev: vnode not a disk"); bo = &vp->v_bufobj; restart: BO_LOCK(bo); TAILQ_FOREACH_SAFE(bp, &bo->bo_dirty.bv_hd, b_bobufs, nbp) { /* * If it is already scheduled, skip to the next buffer. */ if (BUF_LOCK(bp, LK_EXCLUSIVE | LK_NOWAIT, NULL)) continue; if ((bp->b_flags & B_DELWRI) == 0) panic("softdep_fsync_mountdev: not dirty"); /* * We are only interested in bitmaps with outstanding * dependencies. */ if ((wk = LIST_FIRST(&bp->b_dep)) == NULL || wk->wk_type != D_BMSAFEMAP || (bp->b_vflags & BV_BKGRDINPROG)) { BUF_UNLOCK(bp); continue; } BO_UNLOCK(bo); bremfree(bp); (void) bawrite(bp); goto restart; } drain_output(vp); BO_UNLOCK(bo); } /* * Sync all cylinder groups that were dirty at the time this function is * called. Newly dirtied cgs will be inserted before the sentinel. This * is used to flush freedep activity that may be holding up writes to a * indirect block. */ static int sync_cgs(mp, waitfor) struct mount *mp; int waitfor; { struct bmsafemap *bmsafemap; struct bmsafemap *sentinel; struct ufsmount *ump; struct buf *bp; int error; sentinel = malloc(sizeof(*sentinel), M_BMSAFEMAP, M_ZERO | M_WAITOK); sentinel->sm_cg = -1; ump = VFSTOUFS(mp); error = 0; ACQUIRE_LOCK(ump); LIST_INSERT_HEAD(&ump->softdep_dirtycg, sentinel, sm_next); for (bmsafemap = LIST_NEXT(sentinel, sm_next); bmsafemap != NULL; bmsafemap = LIST_NEXT(sentinel, sm_next)) { /* Skip sentinels and cgs with no work to release. */ if (bmsafemap->sm_cg == -1 || (LIST_EMPTY(&bmsafemap->sm_freehd) && LIST_EMPTY(&bmsafemap->sm_freewr))) { LIST_REMOVE(sentinel, sm_next); LIST_INSERT_AFTER(bmsafemap, sentinel, sm_next); continue; } /* * If we don't get the lock and we're waiting try again, if * not move on to the next buf and try to sync it. */ bp = getdirtybuf(bmsafemap->sm_buf, LOCK_PTR(ump), waitfor); if (bp == NULL && waitfor == MNT_WAIT) continue; LIST_REMOVE(sentinel, sm_next); LIST_INSERT_AFTER(bmsafemap, sentinel, sm_next); if (bp == NULL) continue; FREE_LOCK(ump); if (waitfor == MNT_NOWAIT) bawrite(bp); else error = bwrite(bp); ACQUIRE_LOCK(ump); if (error) break; } LIST_REMOVE(sentinel, sm_next); FREE_LOCK(ump); free(sentinel, M_BMSAFEMAP); return (error); } /* * This routine is called when we are trying to synchronously flush a * file. This routine must eliminate any filesystem metadata dependencies * so that the syncing routine can succeed. */ int softdep_sync_metadata(struct vnode *vp) { struct inode *ip; int error; ip = VTOI(vp); KASSERT(MOUNTEDSOFTDEP(UFSTOVFS(ip->i_ump)) != 0, ("softdep_sync_metadata called on non-softdep filesystem")); /* * Ensure that any direct block dependencies have been cleared, * truncations are started, and inode references are journaled. */ ACQUIRE_LOCK(ip->i_ump); /* * Write all journal records to prevent rollbacks on devvp. */ if (vp->v_type == VCHR) softdep_flushjournal(vp->v_mount); error = flush_inodedep_deps(vp, vp->v_mount, ip->i_number); /* * Ensure that all truncates are written so we won't find deps on * indirect blocks. */ process_truncates(vp); FREE_LOCK(ip->i_ump); return (error); } /* * This routine is called when we are attempting to sync a buf with * dependencies. If waitfor is MNT_NOWAIT it attempts to schedule any * other IO it can but returns EBUSY if the buffer is not yet able to * be written. Dependencies which will not cause rollbacks will always * return 0. */ int softdep_sync_buf(struct vnode *vp, struct buf *bp, int waitfor) { struct indirdep *indirdep; struct pagedep *pagedep; struct allocindir *aip; struct newblk *newblk; struct ufsmount *ump; struct buf *nbp; struct worklist *wk; int i, error; KASSERT(MOUNTEDSOFTDEP(vp->v_mount) != 0, ("softdep_sync_buf called on non-softdep filesystem")); /* * For VCHR we just don't want to force flush any dependencies that * will cause rollbacks. */ if (vp->v_type == VCHR) { if (waitfor == MNT_NOWAIT && softdep_count_dependencies(bp, 0)) return (EBUSY); return (0); } ump = VTOI(vp)->i_ump; ACQUIRE_LOCK(ump); /* * As we hold the buffer locked, none of its dependencies * will disappear. */ error = 0; top: LIST_FOREACH(wk, &bp->b_dep, wk_list) { switch (wk->wk_type) { case D_ALLOCDIRECT: case D_ALLOCINDIR: newblk = WK_NEWBLK(wk); if (newblk->nb_jnewblk != NULL) { if (waitfor == MNT_NOWAIT) { error = EBUSY; goto out_unlock; } jwait(&newblk->nb_jnewblk->jn_list, waitfor); goto top; } if (newblk->nb_state & DEPCOMPLETE || waitfor == MNT_NOWAIT) continue; nbp = newblk->nb_bmsafemap->sm_buf; nbp = getdirtybuf(nbp, LOCK_PTR(ump), waitfor); if (nbp == NULL) goto top; FREE_LOCK(ump); if ((error = bwrite(nbp)) != 0) goto out; ACQUIRE_LOCK(ump); continue; case D_INDIRDEP: indirdep = WK_INDIRDEP(wk); if (waitfor == MNT_NOWAIT) { if (!TAILQ_EMPTY(&indirdep->ir_trunc) || !LIST_EMPTY(&indirdep->ir_deplisthd)) { error = EBUSY; goto out_unlock; } } if (!TAILQ_EMPTY(&indirdep->ir_trunc)) panic("softdep_sync_buf: truncation pending."); restart: LIST_FOREACH(aip, &indirdep->ir_deplisthd, ai_next) { newblk = (struct newblk *)aip; if (newblk->nb_jnewblk != NULL) { jwait(&newblk->nb_jnewblk->jn_list, waitfor); goto restart; } if (newblk->nb_state & DEPCOMPLETE) continue; nbp = newblk->nb_bmsafemap->sm_buf; nbp = getdirtybuf(nbp, LOCK_PTR(ump), waitfor); if (nbp == NULL) goto restart; FREE_LOCK(ump); if ((error = bwrite(nbp)) != 0) goto out; ACQUIRE_LOCK(ump); goto restart; } continue; case D_PAGEDEP: /* * Only flush directory entries in synchronous passes. */ if (waitfor != MNT_WAIT) { error = EBUSY; goto out_unlock; } /* * While syncing snapshots, we must allow recursive * lookups. */ BUF_AREC(bp); /* * We are trying to sync a directory that may * have dependencies on both its own metadata * and/or dependencies on the inodes of any * recently allocated files. We walk its diradd * lists pushing out the associated inode. */ pagedep = WK_PAGEDEP(wk); for (i = 0; i < DAHASHSZ; i++) { if (LIST_FIRST(&pagedep->pd_diraddhd[i]) == 0) continue; if ((error = flush_pagedep_deps(vp, wk->wk_mp, &pagedep->pd_diraddhd[i]))) { BUF_NOREC(bp); goto out_unlock; } } BUF_NOREC(bp); continue; case D_FREEWORK: case D_FREEDEP: case D_JSEGDEP: case D_JNEWBLK: continue; default: panic("softdep_sync_buf: Unknown type %s", TYPENAME(wk->wk_type)); /* NOTREACHED */ } } out_unlock: FREE_LOCK(ump); out: return (error); } /* * Flush the dependencies associated with an inodedep. * Called with splbio blocked. */ static int flush_inodedep_deps(vp, mp, ino) struct vnode *vp; struct mount *mp; ino_t ino; { struct inodedep *inodedep; struct inoref *inoref; struct ufsmount *ump; int error, waitfor; /* * This work is done in two passes. The first pass grabs most * of the buffers and begins asynchronously writing them. The * only way to wait for these asynchronous writes is to sleep * on the filesystem vnode which may stay busy for a long time * if the filesystem is active. So, instead, we make a second * pass over the dependencies blocking on each write. In the * usual case we will be blocking against a write that we * initiated, so when it is done the dependency will have been * resolved. Thus the second pass is expected to end quickly. * We give a brief window at the top of the loop to allow * any pending I/O to complete. */ ump = VFSTOUFS(mp); LOCK_OWNED(ump); for (error = 0, waitfor = MNT_NOWAIT; ; ) { if (error) return (error); FREE_LOCK(ump); ACQUIRE_LOCK(ump); restart: if (inodedep_lookup(mp, ino, 0, &inodedep) == 0) return (0); TAILQ_FOREACH(inoref, &inodedep->id_inoreflst, if_deps) { if ((inoref->if_state & (DEPCOMPLETE | GOINGAWAY)) == DEPCOMPLETE) { jwait(&inoref->if_list, MNT_WAIT); goto restart; } } if (flush_deplist(&inodedep->id_inoupdt, waitfor, &error) || flush_deplist(&inodedep->id_newinoupdt, waitfor, &error) || flush_deplist(&inodedep->id_extupdt, waitfor, &error) || flush_deplist(&inodedep->id_newextupdt, waitfor, &error)) continue; /* * If pass2, we are done, otherwise do pass 2. */ if (waitfor == MNT_WAIT) break; waitfor = MNT_WAIT; } /* * Try freeing inodedep in case all dependencies have been removed. */ if (inodedep_lookup(mp, ino, 0, &inodedep) != 0) (void) free_inodedep(inodedep); return (0); } /* * Flush an inode dependency list. * Called with splbio blocked. */ static int flush_deplist(listhead, waitfor, errorp) struct allocdirectlst *listhead; int waitfor; int *errorp; { struct allocdirect *adp; struct newblk *newblk; struct ufsmount *ump; struct buf *bp; if ((adp = TAILQ_FIRST(listhead)) == NULL) return (0); ump = VFSTOUFS(adp->ad_list.wk_mp); LOCK_OWNED(ump); TAILQ_FOREACH(adp, listhead, ad_next) { newblk = (struct newblk *)adp; if (newblk->nb_jnewblk != NULL) { jwait(&newblk->nb_jnewblk->jn_list, MNT_WAIT); return (1); } if (newblk->nb_state & DEPCOMPLETE) continue; bp = newblk->nb_bmsafemap->sm_buf; bp = getdirtybuf(bp, LOCK_PTR(ump), waitfor); if (bp == NULL) { if (waitfor == MNT_NOWAIT) continue; return (1); } FREE_LOCK(ump); if (waitfor == MNT_NOWAIT) bawrite(bp); else *errorp = bwrite(bp); ACQUIRE_LOCK(ump); return (1); } return (0); } /* * Flush dependencies associated with an allocdirect block. */ static int flush_newblk_dep(vp, mp, lbn) struct vnode *vp; struct mount *mp; ufs_lbn_t lbn; { struct newblk *newblk; struct ufsmount *ump; struct bufobj *bo; struct inode *ip; struct buf *bp; ufs2_daddr_t blkno; int error; error = 0; bo = &vp->v_bufobj; ip = VTOI(vp); blkno = DIP(ip, i_db[lbn]); if (blkno == 0) panic("flush_newblk_dep: Missing block"); ump = VFSTOUFS(mp); ACQUIRE_LOCK(ump); /* * Loop until all dependencies related to this block are satisfied. * We must be careful to restart after each sleep in case a write * completes some part of this process for us. */ for (;;) { if (newblk_lookup(mp, blkno, 0, &newblk) == 0) { FREE_LOCK(ump); break; } if (newblk->nb_list.wk_type != D_ALLOCDIRECT) panic("flush_newblk_deps: Bad newblk %p", newblk); /* * Flush the journal. */ if (newblk->nb_jnewblk != NULL) { jwait(&newblk->nb_jnewblk->jn_list, MNT_WAIT); continue; } /* * Write the bitmap dependency. */ if ((newblk->nb_state & DEPCOMPLETE) == 0) { bp = newblk->nb_bmsafemap->sm_buf; bp = getdirtybuf(bp, LOCK_PTR(ump), MNT_WAIT); if (bp == NULL) continue; FREE_LOCK(ump); error = bwrite(bp); if (error) break; ACQUIRE_LOCK(ump); continue; } /* * Write the buffer. */ FREE_LOCK(ump); BO_LOCK(bo); bp = gbincore(bo, lbn); if (bp != NULL) { error = BUF_LOCK(bp, LK_EXCLUSIVE | LK_SLEEPFAIL | LK_INTERLOCK, BO_LOCKPTR(bo)); if (error == ENOLCK) { ACQUIRE_LOCK(ump); continue; /* Slept, retry */ } if (error != 0) break; /* Failed */ if (bp->b_flags & B_DELWRI) { bremfree(bp); error = bwrite(bp); if (error) break; } else BUF_UNLOCK(bp); } else BO_UNLOCK(bo); /* * We have to wait for the direct pointers to * point at the newdirblk before the dependency * will go away. */ error = ffs_update(vp, 1); if (error) break; ACQUIRE_LOCK(ump); } return (error); } /* * Eliminate a pagedep dependency by flushing out all its diradd dependencies. * Called with splbio blocked. */ static int flush_pagedep_deps(pvp, mp, diraddhdp) struct vnode *pvp; struct mount *mp; struct diraddhd *diraddhdp; { struct inodedep *inodedep; struct inoref *inoref; struct ufsmount *ump; struct diradd *dap; struct vnode *vp; int error = 0; struct buf *bp; ino_t inum; struct diraddhd unfinished; LIST_INIT(&unfinished); ump = VFSTOUFS(mp); LOCK_OWNED(ump); restart: while ((dap = LIST_FIRST(diraddhdp)) != NULL) { /* * Flush ourselves if this directory entry * has a MKDIR_PARENT dependency. */ if (dap->da_state & MKDIR_PARENT) { FREE_LOCK(ump); if ((error = ffs_update(pvp, 1)) != 0) break; ACQUIRE_LOCK(ump); /* * If that cleared dependencies, go on to next. */ if (dap != LIST_FIRST(diraddhdp)) continue; /* * All MKDIR_PARENT dependencies and all the * NEWBLOCK pagedeps that are contained in direct * blocks were resolved by doing above ffs_update. * Pagedeps contained in indirect blocks may * require a complete sync'ing of the directory. * We are in the midst of doing a complete sync, * so if they are not resolved in this pass we * defer them for now as they will be sync'ed by * our caller shortly. */ LIST_REMOVE(dap, da_pdlist); LIST_INSERT_HEAD(&unfinished, dap, da_pdlist); continue; } /* * A newly allocated directory must have its "." and * ".." entries written out before its name can be * committed in its parent. */ inum = dap->da_newinum; if (inodedep_lookup(UFSTOVFS(ump), inum, 0, &inodedep) == 0) panic("flush_pagedep_deps: lost inode1"); /* * Wait for any pending journal adds to complete so we don't * cause rollbacks while syncing. */ TAILQ_FOREACH(inoref, &inodedep->id_inoreflst, if_deps) { if ((inoref->if_state & (DEPCOMPLETE | GOINGAWAY)) == DEPCOMPLETE) { jwait(&inoref->if_list, MNT_WAIT); goto restart; } } if (dap->da_state & MKDIR_BODY) { FREE_LOCK(ump); if ((error = ffs_vgetf(mp, inum, LK_EXCLUSIVE, &vp, FFSV_FORCEINSMQ))) break; error = flush_newblk_dep(vp, mp, 0); /* * If we still have the dependency we might need to * update the vnode to sync the new link count to * disk. */ if (error == 0 && dap == LIST_FIRST(diraddhdp)) error = ffs_update(vp, 1); vput(vp); if (error != 0) break; ACQUIRE_LOCK(ump); /* * If that cleared dependencies, go on to next. */ if (dap != LIST_FIRST(diraddhdp)) continue; if (dap->da_state & MKDIR_BODY) { inodedep_lookup(UFSTOVFS(ump), inum, 0, &inodedep); panic("flush_pagedep_deps: MKDIR_BODY " "inodedep %p dap %p vp %p", inodedep, dap, vp); } } /* * Flush the inode on which the directory entry depends. * Having accounted for MKDIR_PARENT and MKDIR_BODY above, * the only remaining dependency is that the updated inode * count must get pushed to disk. The inode has already * been pushed into its inode buffer (via VOP_UPDATE) at * the time of the reference count change. So we need only * locate that buffer, ensure that there will be no rollback * caused by a bitmap dependency, then write the inode buffer. */ retry: if (inodedep_lookup(UFSTOVFS(ump), inum, 0, &inodedep) == 0) panic("flush_pagedep_deps: lost inode"); /* * If the inode still has bitmap dependencies, * push them to disk. */ if ((inodedep->id_state & (DEPCOMPLETE | GOINGAWAY)) == 0) { bp = inodedep->id_bmsafemap->sm_buf; bp = getdirtybuf(bp, LOCK_PTR(ump), MNT_WAIT); if (bp == NULL) goto retry; FREE_LOCK(ump); if ((error = bwrite(bp)) != 0) break; ACQUIRE_LOCK(ump); if (dap != LIST_FIRST(diraddhdp)) continue; } /* * If the inode is still sitting in a buffer waiting * to be written or waiting for the link count to be * adjusted update it here to flush it to disk. */ if (dap == LIST_FIRST(diraddhdp)) { FREE_LOCK(ump); if ((error = ffs_vgetf(mp, inum, LK_EXCLUSIVE, &vp, FFSV_FORCEINSMQ))) break; error = ffs_update(vp, 1); vput(vp); if (error) break; ACQUIRE_LOCK(ump); } /* * If we have failed to get rid of all the dependencies * then something is seriously wrong. */ if (dap == LIST_FIRST(diraddhdp)) { inodedep_lookup(UFSTOVFS(ump), inum, 0, &inodedep); panic("flush_pagedep_deps: failed to flush " "inodedep %p ino %ju dap %p", inodedep, (uintmax_t)inum, dap); } } if (error) ACQUIRE_LOCK(ump); while ((dap = LIST_FIRST(&unfinished)) != NULL) { LIST_REMOVE(dap, da_pdlist); LIST_INSERT_HEAD(diraddhdp, dap, da_pdlist); } return (error); } /* * A large burst of file addition or deletion activity can drive the * memory load excessively high. First attempt to slow things down * using the techniques below. If that fails, this routine requests * the offending operations to fall back to running synchronously * until the memory load returns to a reasonable level. */ int softdep_slowdown(vp) struct vnode *vp; { struct ufsmount *ump; int jlow; int max_softdeps_hard; KASSERT(MOUNTEDSOFTDEP(vp->v_mount) != 0, ("softdep_slowdown called on non-softdep filesystem")); ump = VFSTOUFS(vp->v_mount); ACQUIRE_LOCK(ump); jlow = 0; /* * Check for journal space if needed. */ if (DOINGSUJ(vp)) { if (journal_space(ump, 0) == 0) jlow = 1; } /* * If the system is under its limits and our filesystem is * not responsible for more than our share of the usage and * we are not low on journal space, then no need to slow down. */ max_softdeps_hard = max_softdeps * 11 / 10; if (dep_current[D_DIRREM] < max_softdeps_hard / 2 && dep_current[D_INODEDEP] < max_softdeps_hard && dep_current[D_INDIRDEP] < max_softdeps_hard / 1000 && dep_current[D_FREEBLKS] < max_softdeps_hard && jlow == 0 && ump->softdep_curdeps[D_DIRREM] < (max_softdeps_hard / 2) / stat_flush_threads && ump->softdep_curdeps[D_INODEDEP] < max_softdeps_hard / stat_flush_threads && ump->softdep_curdeps[D_INDIRDEP] < (max_softdeps_hard / 1000) / stat_flush_threads && ump->softdep_curdeps[D_FREEBLKS] < max_softdeps_hard / stat_flush_threads) { FREE_LOCK(ump); return (0); } /* * If the journal is low or our filesystem is over its limit * then speedup the cleanup. */ if (ump->softdep_curdeps[D_INDIRDEP] < (max_softdeps_hard / 1000) / stat_flush_threads || jlow) softdep_speedup(ump); stat_sync_limit_hit += 1; FREE_LOCK(ump); /* * We only slow down the rate at which new dependencies are * generated if we are not using journaling. With journaling, * the cleanup should always be sufficient to keep things * under control. */ if (DOINGSUJ(vp)) return (0); return (1); } /* * Called by the allocation routines when they are about to fail * in the hope that we can free up the requested resource (inodes * or disk space). * * First check to see if the work list has anything on it. If it has, * clean up entries until we successfully free the requested resource. * Because this process holds inodes locked, we cannot handle any remove * requests that might block on a locked inode as that could lead to * deadlock. If the worklist yields none of the requested resource, * start syncing out vnodes to free up the needed space. */ int softdep_request_cleanup(fs, vp, cred, resource) struct fs *fs; struct vnode *vp; struct ucred *cred; int resource; { struct ufsmount *ump; struct mount *mp; struct vnode *lvp, *mvp; long starttime; ufs2_daddr_t needed; int error; /* * If we are being called because of a process doing a * copy-on-write, then it is not safe to process any * worklist items as we will recurse into the copyonwrite * routine. This will result in an incoherent snapshot. * If the vnode that we hold is a snapshot, we must avoid * handling other resources that could cause deadlock. */ if ((curthread->td_pflags & TDP_COWINPROGRESS) || IS_SNAPSHOT(VTOI(vp))) return (0); if (resource == FLUSH_BLOCKS_WAIT) stat_cleanup_blkrequests += 1; else stat_cleanup_inorequests += 1; mp = vp->v_mount; ump = VFSTOUFS(mp); mtx_assert(UFS_MTX(ump), MA_OWNED); UFS_UNLOCK(ump); error = ffs_update(vp, 1); if (error != 0 || MOUNTEDSOFTDEP(mp) == 0) { UFS_LOCK(ump); return (0); } /* * If we are in need of resources, start by cleaning up * any block removals associated with our inode. */ ACQUIRE_LOCK(ump); process_removes(vp); process_truncates(vp); FREE_LOCK(ump); /* * Now clean up at least as many resources as we will need. * * When requested to clean up inodes, the number that are needed * is set by the number of simultaneous writers (mnt_writeopcount) * plus a bit of slop (2) in case some more writers show up while * we are cleaning. * * When requested to free up space, the amount of space that * we need is enough blocks to allocate a full-sized segment * (fs_contigsumsize). The number of such segments that will * be needed is set by the number of simultaneous writers * (mnt_writeopcount) plus a bit of slop (2) in case some more * writers show up while we are cleaning. * * Additionally, if we are unpriviledged and allocating space, * we need to ensure that we clean up enough blocks to get the * needed number of blocks over the threshhold of the minimum * number of blocks required to be kept free by the filesystem * (fs_minfree). */ if (resource == FLUSH_INODES_WAIT) { needed = vp->v_mount->mnt_writeopcount + 2; } else if (resource == FLUSH_BLOCKS_WAIT) { needed = (vp->v_mount->mnt_writeopcount + 2) * fs->fs_contigsumsize; if (priv_check_cred(cred, PRIV_VFS_BLOCKRESERVE, 0)) needed += fragstoblks(fs, roundup((fs->fs_dsize * fs->fs_minfree / 100) - fs->fs_cstotal.cs_nffree, fs->fs_frag)); } else { UFS_LOCK(ump); printf("softdep_request_cleanup: Unknown resource type %d\n", resource); return (0); } starttime = time_second; retry: if ((resource == FLUSH_BLOCKS_WAIT && ump->softdep_on_worklist > 0 && fs->fs_cstotal.cs_nbfree <= needed) || (resource == FLUSH_INODES_WAIT && fs->fs_pendinginodes > 0 && fs->fs_cstotal.cs_nifree <= needed)) { ACQUIRE_LOCK(ump); if (ump->softdep_on_worklist > 0 && process_worklist_item(UFSTOVFS(ump), ump->softdep_on_worklist, LK_NOWAIT) != 0) stat_worklist_push += 1; FREE_LOCK(ump); } /* * If we still need resources and there are no more worklist * entries to process to obtain them, we have to start flushing * the dirty vnodes to force the release of additional requests * to the worklist that we can then process to reap addition * resources. We walk the vnodes associated with the mount point * until we get the needed worklist requests that we can reap. */ if ((resource == FLUSH_BLOCKS_WAIT && fs->fs_cstotal.cs_nbfree <= needed) || (resource == FLUSH_INODES_WAIT && fs->fs_pendinginodes > 0 && fs->fs_cstotal.cs_nifree <= needed)) { MNT_VNODE_FOREACH_ALL(lvp, mp, mvp) { if (TAILQ_FIRST(&lvp->v_bufobj.bo_dirty.bv_hd) == 0) { VI_UNLOCK(lvp); continue; } if (vget(lvp, LK_EXCLUSIVE | LK_INTERLOCK | LK_NOWAIT, curthread)) continue; if (lvp->v_vflag & VV_NOSYNC) { /* unlinked */ vput(lvp); continue; } (void) ffs_syncvnode(lvp, MNT_NOWAIT, 0); vput(lvp); } lvp = ump->um_devvp; if (vn_lock(lvp, LK_EXCLUSIVE | LK_NOWAIT) == 0) { VOP_FSYNC(lvp, MNT_NOWAIT, curthread); VOP_UNLOCK(lvp, 0); } if (ump->softdep_on_worklist > 0) { stat_cleanup_retries += 1; goto retry; } stat_cleanup_failures += 1; } if (time_second - starttime > stat_cleanup_high_delay) stat_cleanup_high_delay = time_second - starttime; UFS_LOCK(ump); return (1); } static bool softdep_excess_inodes(struct ufsmount *ump) { return (dep_current[D_INODEDEP] > max_softdeps && ump->softdep_curdeps[D_INODEDEP] > max_softdeps / stat_flush_threads); } static bool softdep_excess_dirrem(struct ufsmount *ump) { return (dep_current[D_DIRREM] > max_softdeps / 2 && ump->softdep_curdeps[D_DIRREM] > (max_softdeps / 2) / stat_flush_threads); } static void schedule_cleanup(struct mount *mp) { struct ufsmount *ump; struct thread *td; ump = VFSTOUFS(mp); LOCK_OWNED(ump); FREE_LOCK(ump); td = curthread; if ((td->td_pflags & TDP_KTHREAD) != 0 && (td->td_proc->p_flag2 & P2_AST_SU) == 0) { /* * No ast is delivered to kernel threads, so nobody * would deref the mp. Some kernel threads * explicitely check for AST, e.g. NFS daemon does * this in the serving loop. */ return; } if (td->td_su != NULL) vfs_rel(td->td_su); vfs_ref(mp); td->td_su = mp; thread_lock(td); td->td_flags |= TDF_ASTPENDING; thread_unlock(td); } static void softdep_ast_cleanup_proc(void) { struct thread *td; struct mount *mp; struct ufsmount *ump; int error; bool req; td = curthread; mp = td->td_su; if (mp == NULL) return; td->td_su = NULL; error = vfs_busy(mp, MBF_NOWAIT); vfs_rel(mp); if (error != 0) return; if (ffs_own_mount(mp) && MOUNTEDSOFTDEP(mp)) { ump = VFSTOUFS(mp); for (;;) { req = false; ACQUIRE_LOCK(ump); if (softdep_excess_inodes(ump)) { req = true; request_cleanup(mp, FLUSH_INODES); } if (softdep_excess_dirrem(ump)) { req = true; request_cleanup(mp, FLUSH_BLOCKS); } FREE_LOCK(ump); if ((td->td_pflags & TDP_KTHREAD) != 0 || !req) break; } } vfs_unbusy(mp); } /* * If memory utilization has gotten too high, deliberately slow things * down and speed up the I/O processing. */ static int request_cleanup(mp, resource) struct mount *mp; int resource; { struct thread *td = curthread; struct ufsmount *ump; ump = VFSTOUFS(mp); LOCK_OWNED(ump); /* * We never hold up the filesystem syncer or buf daemon. */ if (td->td_pflags & (TDP_SOFTDEP|TDP_NORUNNINGBUF)) return (0); /* * First check to see if the work list has gotten backlogged. * If it has, co-opt this process to help clean up two entries. * Because this process may hold inodes locked, we cannot * handle any remove requests that might block on a locked * inode as that could lead to deadlock. We set TDP_SOFTDEP * to avoid recursively processing the worklist. */ if (ump->softdep_on_worklist > max_softdeps / 10) { td->td_pflags |= TDP_SOFTDEP; process_worklist_item(mp, 2, LK_NOWAIT); td->td_pflags &= ~TDP_SOFTDEP; stat_worklist_push += 2; return(1); } /* * Next, we attempt to speed up the syncer process. If that * is successful, then we allow the process to continue. */ if (softdep_speedup(ump) && resource != FLUSH_BLOCKS_WAIT && resource != FLUSH_INODES_WAIT) return(0); /* * If we are resource constrained on inode dependencies, try * flushing some dirty inodes. Otherwise, we are constrained * by file deletions, so try accelerating flushes of directories * with removal dependencies. We would like to do the cleanup * here, but we probably hold an inode locked at this point and * that might deadlock against one that we try to clean. So, * the best that we can do is request the syncer daemon to do * the cleanup for us. */ switch (resource) { case FLUSH_INODES: case FLUSH_INODES_WAIT: ACQUIRE_GBLLOCK(&lk); stat_ino_limit_push += 1; req_clear_inodedeps += 1; FREE_GBLLOCK(&lk); stat_countp = &stat_ino_limit_hit; break; case FLUSH_BLOCKS: case FLUSH_BLOCKS_WAIT: ACQUIRE_GBLLOCK(&lk); stat_blk_limit_push += 1; req_clear_remove += 1; FREE_GBLLOCK(&lk); stat_countp = &stat_blk_limit_hit; break; default: panic("request_cleanup: unknown type"); } /* * Hopefully the syncer daemon will catch up and awaken us. * We wait at most tickdelay before proceeding in any case. */ ACQUIRE_GBLLOCK(&lk); FREE_LOCK(ump); proc_waiting += 1; if (callout_pending(&softdep_callout) == FALSE) callout_reset(&softdep_callout, tickdelay > 2 ? tickdelay : 2, pause_timer, 0); if ((td->td_pflags & TDP_KTHREAD) == 0) msleep((caddr_t)&proc_waiting, &lk, PPAUSE, "softupdate", 0); proc_waiting -= 1; FREE_GBLLOCK(&lk); ACQUIRE_LOCK(ump); return (1); } /* * Awaken processes pausing in request_cleanup and clear proc_waiting * to indicate that there is no longer a timer running. Pause_timer * will be called with the global softdep mutex (&lk) locked. */ static void pause_timer(arg) void *arg; { GBLLOCK_OWNED(&lk); /* * The callout_ API has acquired mtx and will hold it around this * function call. */ *stat_countp += proc_waiting; wakeup(&proc_waiting); } /* * If requested, try removing inode or removal dependencies. */ static void check_clear_deps(mp) struct mount *mp; { /* * If we are suspended, it may be because of our using * too many inodedeps, so help clear them out. */ if (MOUNTEDSUJ(mp) && VFSTOUFS(mp)->softdep_jblocks->jb_suspended) clear_inodedeps(mp); /* * General requests for cleanup of backed up dependencies */ ACQUIRE_GBLLOCK(&lk); if (req_clear_inodedeps) { req_clear_inodedeps -= 1; FREE_GBLLOCK(&lk); clear_inodedeps(mp); ACQUIRE_GBLLOCK(&lk); wakeup(&proc_waiting); } if (req_clear_remove) { req_clear_remove -= 1; FREE_GBLLOCK(&lk); clear_remove(mp); ACQUIRE_GBLLOCK(&lk); wakeup(&proc_waiting); } FREE_GBLLOCK(&lk); } /* * Flush out a directory with at least one removal dependency in an effort to * reduce the number of dirrem, freefile, and freeblks dependency structures. */ static void clear_remove(mp) struct mount *mp; { struct pagedep_hashhead *pagedephd; struct pagedep *pagedep; struct ufsmount *ump; struct vnode *vp; struct bufobj *bo; int error, cnt; ino_t ino; ump = VFSTOUFS(mp); LOCK_OWNED(ump); for (cnt = 0; cnt <= ump->pagedep_hash_size; cnt++) { pagedephd = &ump->pagedep_hashtbl[ump->pagedep_nextclean++]; if (ump->pagedep_nextclean > ump->pagedep_hash_size) ump->pagedep_nextclean = 0; LIST_FOREACH(pagedep, pagedephd, pd_hash) { if (LIST_EMPTY(&pagedep->pd_dirremhd)) continue; ino = pagedep->pd_ino; if (vn_start_write(NULL, &mp, V_NOWAIT) != 0) continue; FREE_LOCK(ump); /* * Let unmount clear deps */ error = vfs_busy(mp, MBF_NOWAIT); if (error != 0) goto finish_write; error = ffs_vgetf(mp, ino, LK_EXCLUSIVE, &vp, FFSV_FORCEINSMQ); vfs_unbusy(mp); if (error != 0) { softdep_error("clear_remove: vget", error); goto finish_write; } if ((error = ffs_syncvnode(vp, MNT_NOWAIT, 0))) softdep_error("clear_remove: fsync", error); bo = &vp->v_bufobj; BO_LOCK(bo); drain_output(vp); BO_UNLOCK(bo); vput(vp); finish_write: vn_finished_write(mp); ACQUIRE_LOCK(ump); return; } } } /* * Clear out a block of dirty inodes in an effort to reduce * the number of inodedep dependency structures. */ static void clear_inodedeps(mp) struct mount *mp; { struct inodedep_hashhead *inodedephd; struct inodedep *inodedep; struct ufsmount *ump; struct vnode *vp; struct fs *fs; int error, cnt; ino_t firstino, lastino, ino; ump = VFSTOUFS(mp); fs = ump->um_fs; LOCK_OWNED(ump); /* * Pick a random inode dependency to be cleared. * We will then gather up all the inodes in its block * that have dependencies and flush them out. */ for (cnt = 0; cnt <= ump->inodedep_hash_size; cnt++) { inodedephd = &ump->inodedep_hashtbl[ump->inodedep_nextclean++]; if (ump->inodedep_nextclean > ump->inodedep_hash_size) ump->inodedep_nextclean = 0; if ((inodedep = LIST_FIRST(inodedephd)) != NULL) break; } if (inodedep == NULL) return; /* * Find the last inode in the block with dependencies. */ firstino = inodedep->id_ino & ~(INOPB(fs) - 1); for (lastino = firstino + INOPB(fs) - 1; lastino > firstino; lastino--) if (inodedep_lookup(mp, lastino, 0, &inodedep) != 0) break; /* * Asynchronously push all but the last inode with dependencies. * Synchronously push the last inode with dependencies to ensure * that the inode block gets written to free up the inodedeps. */ for (ino = firstino; ino <= lastino; ino++) { if (inodedep_lookup(mp, ino, 0, &inodedep) == 0) continue; if (vn_start_write(NULL, &mp, V_NOWAIT) != 0) continue; FREE_LOCK(ump); error = vfs_busy(mp, MBF_NOWAIT); /* Let unmount clear deps */ if (error != 0) { vn_finished_write(mp); ACQUIRE_LOCK(ump); return; } if ((error = ffs_vgetf(mp, ino, LK_EXCLUSIVE, &vp, FFSV_FORCEINSMQ)) != 0) { softdep_error("clear_inodedeps: vget", error); vfs_unbusy(mp); vn_finished_write(mp); ACQUIRE_LOCK(ump); return; } vfs_unbusy(mp); if (ino == lastino) { if ((error = ffs_syncvnode(vp, MNT_WAIT, 0))) softdep_error("clear_inodedeps: fsync1", error); } else { if ((error = ffs_syncvnode(vp, MNT_NOWAIT, 0))) softdep_error("clear_inodedeps: fsync2", error); BO_LOCK(&vp->v_bufobj); drain_output(vp); BO_UNLOCK(&vp->v_bufobj); } vput(vp); vn_finished_write(mp); ACQUIRE_LOCK(ump); } } void softdep_buf_append(bp, wkhd) struct buf *bp; struct workhead *wkhd; { struct worklist *wk; struct ufsmount *ump; if ((wk = LIST_FIRST(wkhd)) == NULL) return; KASSERT(MOUNTEDSOFTDEP(wk->wk_mp) != 0, ("softdep_buf_append called on non-softdep filesystem")); ump = VFSTOUFS(wk->wk_mp); ACQUIRE_LOCK(ump); while ((wk = LIST_FIRST(wkhd)) != NULL) { WORKLIST_REMOVE(wk); WORKLIST_INSERT(&bp->b_dep, wk); } FREE_LOCK(ump); } void softdep_inode_append(ip, cred, wkhd) struct inode *ip; struct ucred *cred; struct workhead *wkhd; { struct buf *bp; struct fs *fs; int error; KASSERT(MOUNTEDSOFTDEP(UFSTOVFS(ip->i_ump)) != 0, ("softdep_inode_append called on non-softdep filesystem")); fs = ip->i_fs; error = bread(ip->i_devvp, fsbtodb(fs, ino_to_fsba(fs, ip->i_number)), (int)fs->fs_bsize, cred, &bp); if (error) { bqrelse(bp); softdep_freework(wkhd); return; } softdep_buf_append(bp, wkhd); bqrelse(bp); } void softdep_freework(wkhd) struct workhead *wkhd; { struct worklist *wk; struct ufsmount *ump; if ((wk = LIST_FIRST(wkhd)) == NULL) return; KASSERT(MOUNTEDSOFTDEP(wk->wk_mp) != 0, ("softdep_freework called on non-softdep filesystem")); ump = VFSTOUFS(wk->wk_mp); ACQUIRE_LOCK(ump); handle_jwork(wkhd); FREE_LOCK(ump); } /* * Function to determine if the buffer has outstanding dependencies * that will cause a roll-back if the buffer is written. If wantcount * is set, return number of dependencies, otherwise just yes or no. */ static int softdep_count_dependencies(bp, wantcount) struct buf *bp; int wantcount; { struct worklist *wk; struct ufsmount *ump; struct bmsafemap *bmsafemap; struct freework *freework; struct inodedep *inodedep; struct indirdep *indirdep; struct freeblks *freeblks; struct allocindir *aip; struct pagedep *pagedep; struct dirrem *dirrem; struct newblk *newblk; struct mkdir *mkdir; struct diradd *dap; int i, retval; retval = 0; if ((wk = LIST_FIRST(&bp->b_dep)) == NULL) return (0); ump = VFSTOUFS(wk->wk_mp); ACQUIRE_LOCK(ump); LIST_FOREACH(wk, &bp->b_dep, wk_list) { switch (wk->wk_type) { case D_INODEDEP: inodedep = WK_INODEDEP(wk); if ((inodedep->id_state & DEPCOMPLETE) == 0) { /* bitmap allocation dependency */ retval += 1; if (!wantcount) goto out; } if (TAILQ_FIRST(&inodedep->id_inoupdt)) { /* direct block pointer dependency */ retval += 1; if (!wantcount) goto out; } if (TAILQ_FIRST(&inodedep->id_extupdt)) { /* direct block pointer dependency */ retval += 1; if (!wantcount) goto out; } if (TAILQ_FIRST(&inodedep->id_inoreflst)) { /* Add reference dependency. */ retval += 1; if (!wantcount) goto out; } continue; case D_INDIRDEP: indirdep = WK_INDIRDEP(wk); TAILQ_FOREACH(freework, &indirdep->ir_trunc, fw_next) { /* indirect truncation dependency */ retval += 1; if (!wantcount) goto out; } LIST_FOREACH(aip, &indirdep->ir_deplisthd, ai_next) { /* indirect block pointer dependency */ retval += 1; if (!wantcount) goto out; } continue; case D_PAGEDEP: pagedep = WK_PAGEDEP(wk); LIST_FOREACH(dirrem, &pagedep->pd_dirremhd, dm_next) { if (LIST_FIRST(&dirrem->dm_jremrefhd)) { /* Journal remove ref dependency. */ retval += 1; if (!wantcount) goto out; } } for (i = 0; i < DAHASHSZ; i++) { LIST_FOREACH(dap, &pagedep->pd_diraddhd[i], da_pdlist) { /* directory entry dependency */ retval += 1; if (!wantcount) goto out; } } continue; case D_BMSAFEMAP: bmsafemap = WK_BMSAFEMAP(wk); if (LIST_FIRST(&bmsafemap->sm_jaddrefhd)) { /* Add reference dependency. */ retval += 1; if (!wantcount) goto out; } if (LIST_FIRST(&bmsafemap->sm_jnewblkhd)) { /* Allocate block dependency. */ retval += 1; if (!wantcount) goto out; } continue; case D_FREEBLKS: freeblks = WK_FREEBLKS(wk); if (LIST_FIRST(&freeblks->fb_jblkdephd)) { /* Freeblk journal dependency. */ retval += 1; if (!wantcount) goto out; } continue; case D_ALLOCDIRECT: case D_ALLOCINDIR: newblk = WK_NEWBLK(wk); if (newblk->nb_jnewblk) { /* Journal allocate dependency. */ retval += 1; if (!wantcount) goto out; } continue; case D_MKDIR: mkdir = WK_MKDIR(wk); if (mkdir->md_jaddref) { /* Journal reference dependency. */ retval += 1; if (!wantcount) goto out; } continue; case D_FREEWORK: case D_FREEDEP: case D_JSEGDEP: case D_JSEG: case D_SBDEP: /* never a dependency on these blocks */ continue; default: panic("softdep_count_dependencies: Unexpected type %s", TYPENAME(wk->wk_type)); /* NOTREACHED */ } } out: FREE_LOCK(ump); return retval; } /* * Acquire exclusive access to a buffer. * Must be called with a locked mtx parameter. * Return acquired buffer or NULL on failure. */ static struct buf * getdirtybuf(bp, lock, waitfor) struct buf *bp; struct rwlock *lock; int waitfor; { int error; if (BUF_LOCK(bp, LK_EXCLUSIVE | LK_NOWAIT, NULL) != 0) { if (waitfor != MNT_WAIT) return (NULL); error = BUF_LOCK(bp, LK_EXCLUSIVE | LK_SLEEPFAIL | LK_INTERLOCK, lock); /* * Even if we sucessfully acquire bp here, we have dropped * lock, which may violates our guarantee. */ if (error == 0) BUF_UNLOCK(bp); else if (error != ENOLCK) panic("getdirtybuf: inconsistent lock: %d", error); rw_wlock(lock); return (NULL); } if ((bp->b_vflags & BV_BKGRDINPROG) != 0) { if (lock != BO_LOCKPTR(bp->b_bufobj) && waitfor == MNT_WAIT) { rw_wunlock(lock); BO_LOCK(bp->b_bufobj); BUF_UNLOCK(bp); if ((bp->b_vflags & BV_BKGRDINPROG) != 0) { bp->b_vflags |= BV_BKGRDWAIT; msleep(&bp->b_xflags, BO_LOCKPTR(bp->b_bufobj), PRIBIO | PDROP, "getbuf", 0); } else BO_UNLOCK(bp->b_bufobj); rw_wlock(lock); return (NULL); } BUF_UNLOCK(bp); if (waitfor != MNT_WAIT) return (NULL); /* * The lock argument must be bp->b_vp's mutex in * this case. */ #ifdef DEBUG_VFS_LOCKS if (bp->b_vp->v_type != VCHR) ASSERT_BO_WLOCKED(bp->b_bufobj); #endif bp->b_vflags |= BV_BKGRDWAIT; rw_sleep(&bp->b_xflags, lock, PRIBIO, "getbuf", 0); return (NULL); } if ((bp->b_flags & B_DELWRI) == 0) { BUF_UNLOCK(bp); return (NULL); } bremfree(bp); return (bp); } /* * Check if it is safe to suspend the file system now. On entry, * the vnode interlock for devvp should be held. Return 0 with * the mount interlock held if the file system can be suspended now, * otherwise return EAGAIN with the mount interlock held. */ int softdep_check_suspend(struct mount *mp, struct vnode *devvp, int softdep_depcnt, int softdep_accdepcnt, int secondary_writes, int secondary_accwrites) { struct bufobj *bo; struct ufsmount *ump; struct inodedep *inodedep; int error, unlinked; bo = &devvp->v_bufobj; ASSERT_BO_WLOCKED(bo); /* * If we are not running with soft updates, then we need only * deal with secondary writes as we try to suspend. */ if (MOUNTEDSOFTDEP(mp) == 0) { MNT_ILOCK(mp); while (mp->mnt_secondary_writes != 0) { BO_UNLOCK(bo); msleep(&mp->mnt_secondary_writes, MNT_MTX(mp), (PUSER - 1) | PDROP, "secwr", 0); BO_LOCK(bo); MNT_ILOCK(mp); } /* * Reasons for needing more work before suspend: * - Dirty buffers on devvp. * - Secondary writes occurred after start of vnode sync loop */ error = 0; if (bo->bo_numoutput > 0 || bo->bo_dirty.bv_cnt > 0 || secondary_writes != 0 || mp->mnt_secondary_writes != 0 || secondary_accwrites != mp->mnt_secondary_accwrites) error = EAGAIN; BO_UNLOCK(bo); return (error); } /* * If we are running with soft updates, then we need to coordinate * with them as we try to suspend. */ ump = VFSTOUFS(mp); for (;;) { if (!TRY_ACQUIRE_LOCK(ump)) { BO_UNLOCK(bo); ACQUIRE_LOCK(ump); FREE_LOCK(ump); BO_LOCK(bo); continue; } MNT_ILOCK(mp); if (mp->mnt_secondary_writes != 0) { FREE_LOCK(ump); BO_UNLOCK(bo); msleep(&mp->mnt_secondary_writes, MNT_MTX(mp), (PUSER - 1) | PDROP, "secwr", 0); BO_LOCK(bo); continue; } break; } unlinked = 0; if (MOUNTEDSUJ(mp)) { for (inodedep = TAILQ_FIRST(&ump->softdep_unlinked); inodedep != NULL; inodedep = TAILQ_NEXT(inodedep, id_unlinked)) { if ((inodedep->id_state & (UNLINKED | UNLINKLINKS | UNLINKONLIST)) != (UNLINKED | UNLINKLINKS | UNLINKONLIST) || !check_inodedep_free(inodedep)) continue; unlinked++; } } /* * Reasons for needing more work before suspend: * - Dirty buffers on devvp. * - Softdep activity occurred after start of vnode sync loop * - Secondary writes occurred after start of vnode sync loop */ error = 0; if (bo->bo_numoutput > 0 || bo->bo_dirty.bv_cnt > 0 || softdep_depcnt != unlinked || ump->softdep_deps != unlinked || softdep_accdepcnt != ump->softdep_accdeps || secondary_writes != 0 || mp->mnt_secondary_writes != 0 || secondary_accwrites != mp->mnt_secondary_accwrites) error = EAGAIN; FREE_LOCK(ump); BO_UNLOCK(bo); return (error); } /* * Get the number of dependency structures for the file system, both * the current number and the total number allocated. These will * later be used to detect that softdep processing has occurred. */ void softdep_get_depcounts(struct mount *mp, int *softdep_depsp, int *softdep_accdepsp) { struct ufsmount *ump; if (MOUNTEDSOFTDEP(mp) == 0) { *softdep_depsp = 0; *softdep_accdepsp = 0; return; } ump = VFSTOUFS(mp); ACQUIRE_LOCK(ump); *softdep_depsp = ump->softdep_deps; *softdep_accdepsp = ump->softdep_accdeps; FREE_LOCK(ump); } /* * Wait for pending output on a vnode to complete. * Must be called with vnode lock and interlock locked. * * XXX: Should just be a call to bufobj_wwait(). */ static void drain_output(vp) struct vnode *vp; { struct bufobj *bo; bo = &vp->v_bufobj; ASSERT_VOP_LOCKED(vp, "drain_output"); ASSERT_BO_WLOCKED(bo); while (bo->bo_numoutput) { bo->bo_flag |= BO_WWAIT; msleep((caddr_t)&bo->bo_numoutput, BO_LOCKPTR(bo), PRIBIO + 1, "drainvp", 0); } } /* * Called whenever a buffer that is being invalidated or reallocated * contains dependencies. This should only happen if an I/O error has * occurred. The routine is called with the buffer locked. */ static void softdep_deallocate_dependencies(bp) struct buf *bp; { if ((bp->b_ioflags & BIO_ERROR) == 0) panic("softdep_deallocate_dependencies: dangling deps"); if (bp->b_vp != NULL && bp->b_vp->v_mount != NULL) softdep_error(bp->b_vp->v_mount->mnt_stat.f_mntonname, bp->b_error); else printf("softdep_deallocate_dependencies: " "got error %d while accessing filesystem\n", bp->b_error); if (bp->b_error != ENXIO) panic("softdep_deallocate_dependencies: unrecovered I/O error"); } /* * Function to handle asynchronous write errors in the filesystem. */ static void softdep_error(func, error) char *func; int error; { /* XXX should do something better! */ printf("%s: got error %d while accessing filesystem\n", func, error); } #ifdef DDB static void inodedep_print(struct inodedep *inodedep, int verbose) { db_printf("%p fs %p st %x ino %jd inoblk %jd delta %d nlink %d" " saveino %p\n", inodedep, inodedep->id_fs, inodedep->id_state, (intmax_t)inodedep->id_ino, (intmax_t)fsbtodb(inodedep->id_fs, ino_to_fsba(inodedep->id_fs, inodedep->id_ino)), inodedep->id_nlinkdelta, inodedep->id_savednlink, inodedep->id_savedino1); if (verbose == 0) return; db_printf("\tpendinghd %p, bufwait %p, inowait %p, inoreflst %p, " "mkdiradd %p\n", LIST_FIRST(&inodedep->id_pendinghd), LIST_FIRST(&inodedep->id_bufwait), LIST_FIRST(&inodedep->id_inowait), TAILQ_FIRST(&inodedep->id_inoreflst), inodedep->id_mkdiradd); db_printf("\tinoupdt %p, newinoupdt %p, extupdt %p, newextupdt %p\n", TAILQ_FIRST(&inodedep->id_inoupdt), TAILQ_FIRST(&inodedep->id_newinoupdt), TAILQ_FIRST(&inodedep->id_extupdt), TAILQ_FIRST(&inodedep->id_newextupdt)); } DB_SHOW_COMMAND(inodedep, db_show_inodedep) { if (have_addr == 0) { db_printf("Address required\n"); return; } inodedep_print((struct inodedep*)addr, 1); } DB_SHOW_COMMAND(inodedeps, db_show_inodedeps) { struct inodedep_hashhead *inodedephd; struct inodedep *inodedep; struct ufsmount *ump; int cnt; if (have_addr == 0) { db_printf("Address required\n"); return; } ump = (struct ufsmount *)addr; for (cnt = 0; cnt < ump->inodedep_hash_size; cnt++) { inodedephd = &ump->inodedep_hashtbl[cnt]; LIST_FOREACH(inodedep, inodedephd, id_hash) { inodedep_print(inodedep, 0); } } } DB_SHOW_COMMAND(worklist, db_show_worklist) { struct worklist *wk; if (have_addr == 0) { db_printf("Address required\n"); return; } wk = (struct worklist *)addr; printf("worklist: %p type %s state 0x%X\n", wk, TYPENAME(wk->wk_type), wk->wk_state); } DB_SHOW_COMMAND(workhead, db_show_workhead) { struct workhead *wkhd; struct worklist *wk; int i; if (have_addr == 0) { db_printf("Address required\n"); return; } wkhd = (struct workhead *)addr; wk = LIST_FIRST(wkhd); for (i = 0; i < 100 && wk != NULL; i++, wk = LIST_NEXT(wk, wk_list)) db_printf("worklist: %p type %s state 0x%X", wk, TYPENAME(wk->wk_type), wk->wk_state); if (i == 100) db_printf("workhead overflow"); printf("\n"); } DB_SHOW_COMMAND(mkdirs, db_show_mkdirs) { struct mkdirlist *mkdirlisthd; struct jaddref *jaddref; struct diradd *diradd; struct mkdir *mkdir; if (have_addr == 0) { db_printf("Address required\n"); return; } mkdirlisthd = (struct mkdirlist *)addr; LIST_FOREACH(mkdir, mkdirlisthd, md_mkdirs) { diradd = mkdir->md_diradd; db_printf("mkdir: %p state 0x%X dap %p state 0x%X", mkdir, mkdir->md_state, diradd, diradd->da_state); if ((jaddref = mkdir->md_jaddref) != NULL) db_printf(" jaddref %p jaddref state 0x%X", jaddref, jaddref->ja_state); db_printf("\n"); } } /* exported to ffs_vfsops.c */ extern void db_print_ffs(struct ufsmount *ump); void db_print_ffs(struct ufsmount *ump) { db_printf("mp %p %s devvp %p fs %p su_wl %d su_deps %d su_req %d\n", ump->um_mountp, ump->um_mountp->mnt_stat.f_mntonname, ump->um_devvp, ump->um_fs, ump->softdep_on_worklist, ump->softdep_deps, ump->softdep_req); } #endif /* DDB */ #endif /* SOFTUPDATES */ Index: projects/clang-trunk/sys/vm/vm_page.c =================================================================== --- projects/clang-trunk/sys/vm/vm_page.c (revision 283870) +++ projects/clang-trunk/sys/vm/vm_page.c (revision 283871) @@ -1,3366 +1,3367 @@ /*- * Copyright (c) 1991 Regents of the University of California. * All rights reserved. * Copyright (c) 1998 Matthew Dillon. All Rights Reserved. * * This code is derived from software contributed to Berkeley by * The Mach Operating System project at Carnegie-Mellon University. * * Redistribution and use in source and binary forms, with or without * modification, are permitted provided that the following conditions * are met: * 1. Redistributions of source code must retain the above copyright * notice, this list of conditions and the following disclaimer. * 2. Redistributions in binary form must reproduce the above copyright * notice, this list of conditions and the following disclaimer in the * documentation and/or other materials provided with the distribution. * 4. Neither the name of the University nor the names of its contributors * may be used to endorse or promote products derived from this software * without specific prior written permission. * * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF * SUCH DAMAGE. * * from: @(#)vm_page.c 7.4 (Berkeley) 5/7/91 */ /*- * Copyright (c) 1987, 1990 Carnegie-Mellon University. * All rights reserved. * * Authors: Avadis Tevanian, Jr., Michael Wayne Young * * Permission to use, copy, modify and distribute this software and * its documentation is hereby granted, provided that both the copyright * notice and this permission notice appear in all copies of the * software, derivative works or modified versions, and any portions * thereof, and that both notices appear in supporting documentation. * * CARNEGIE MELLON ALLOWS FREE USE OF THIS SOFTWARE IN ITS "AS IS" * CONDITION. CARNEGIE MELLON DISCLAIMS ANY LIABILITY OF ANY KIND * FOR ANY DAMAGES WHATSOEVER RESULTING FROM THE USE OF THIS SOFTWARE. * * Carnegie Mellon requests users of this software to return to * * Software Distribution Coordinator or Software.Distribution@CS.CMU.EDU * School of Computer Science * Carnegie Mellon University * Pittsburgh PA 15213-3890 * * any improvements or extensions that they make and grant Carnegie the * rights to redistribute these changes. */ /* * GENERAL RULES ON VM_PAGE MANIPULATION * * - A page queue lock is required when adding or removing a page from a * page queue regardless of other locks or the busy state of a page. * * * In general, no thread besides the page daemon can acquire or * hold more than one page queue lock at a time. * * * The page daemon can acquire and hold any pair of page queue * locks in any order. * * - The object lock is required when inserting or removing * pages from an object (vm_page_insert() or vm_page_remove()). * */ /* * Resident memory management module. */ #include __FBSDID("$FreeBSD$"); #include "opt_vm.h" #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include /* * Associated with page of user-allocatable memory is a * page structure. */ struct vm_domain vm_dom[MAXMEMDOM]; struct mtx_padalign vm_page_queue_free_mtx; struct mtx_padalign pa_lock[PA_LOCK_COUNT]; vm_page_t vm_page_array; long vm_page_array_size; long first_page; int vm_page_zero_count; static int boot_pages = UMA_BOOT_PAGES; SYSCTL_INT(_vm, OID_AUTO, boot_pages, CTLFLAG_RDTUN | CTLFLAG_NOFETCH, &boot_pages, 0, "number of pages allocated for bootstrapping the VM system"); static int pa_tryrelock_restart; SYSCTL_INT(_vm, OID_AUTO, tryrelock_restart, CTLFLAG_RD, &pa_tryrelock_restart, 0, "Number of tryrelock restarts"); static TAILQ_HEAD(, vm_page) blacklist_head; static int sysctl_vm_page_blacklist(SYSCTL_HANDLER_ARGS); SYSCTL_PROC(_vm, OID_AUTO, page_blacklist, CTLTYPE_STRING | CTLFLAG_RD | CTLFLAG_MPSAFE, NULL, 0, sysctl_vm_page_blacklist, "A", "Blacklist pages"); static uma_zone_t fakepg_zone; static struct vnode *vm_page_alloc_init(vm_page_t m); static void vm_page_cache_turn_free(vm_page_t m); static void vm_page_clear_dirty_mask(vm_page_t m, vm_page_bits_t pagebits); static void vm_page_enqueue(uint8_t queue, vm_page_t m); static void vm_page_init_fakepg(void *dummy); static int vm_page_insert_after(vm_page_t m, vm_object_t object, vm_pindex_t pindex, vm_page_t mpred); static void vm_page_insert_radixdone(vm_page_t m, vm_object_t object, vm_page_t mpred); SYSINIT(vm_page, SI_SUB_VM, SI_ORDER_SECOND, vm_page_init_fakepg, NULL); static void vm_page_init_fakepg(void *dummy) { fakepg_zone = uma_zcreate("fakepg", sizeof(struct vm_page), NULL, NULL, NULL, NULL, UMA_ALIGN_PTR, UMA_ZONE_NOFREE | UMA_ZONE_VM); } /* Make sure that u_long is at least 64 bits when PAGE_SIZE is 32K. */ #if PAGE_SIZE == 32768 #ifdef CTASSERT CTASSERT(sizeof(u_long) >= 8); #endif #endif /* * Try to acquire a physical address lock while a pmap is locked. If we * fail to trylock we unlock and lock the pmap directly and cache the * locked pa in *locked. The caller should then restart their loop in case * the virtual to physical mapping has changed. */ int vm_page_pa_tryrelock(pmap_t pmap, vm_paddr_t pa, vm_paddr_t *locked) { vm_paddr_t lockpa; lockpa = *locked; *locked = pa; if (lockpa) { PA_LOCK_ASSERT(lockpa, MA_OWNED); if (PA_LOCKPTR(pa) == PA_LOCKPTR(lockpa)) return (0); PA_UNLOCK(lockpa); } if (PA_TRYLOCK(pa)) return (0); PMAP_UNLOCK(pmap); atomic_add_int(&pa_tryrelock_restart, 1); PA_LOCK(pa); PMAP_LOCK(pmap); return (EAGAIN); } /* * vm_set_page_size: * * Sets the page size, perhaps based upon the memory * size. Must be called before any use of page-size * dependent functions. */ void vm_set_page_size(void) { if (vm_cnt.v_page_size == 0) vm_cnt.v_page_size = PAGE_SIZE; if (((vm_cnt.v_page_size - 1) & vm_cnt.v_page_size) != 0) panic("vm_set_page_size: page size not a power of two"); } /* * vm_page_blacklist_next: * * Find the next entry in the provided string of blacklist * addresses. Entries are separated by space, comma, or newline. * If an invalid integer is encountered then the rest of the * string is skipped. Updates the list pointer to the next * character, or NULL if the string is exhausted or invalid. */ static vm_paddr_t vm_page_blacklist_next(char **list, char *end) { vm_paddr_t bad; char *cp, *pos; if (list == NULL || *list == NULL) return (0); if (**list =='\0') { *list = NULL; return (0); } /* * If there's no end pointer then the buffer is coming from * the kenv and we know it's null-terminated. */ if (end == NULL) end = *list + strlen(*list); /* Ensure that strtoq() won't walk off the end */ if (*end != '\0') { if (*end == '\n' || *end == ' ' || *end == ',') *end = '\0'; else { printf("Blacklist not terminated, skipping\n"); *list = NULL; return (0); } } for (pos = *list; *pos != '\0'; pos = cp) { bad = strtoq(pos, &cp, 0); if (*cp == '\0' || *cp == ' ' || *cp == ',' || *cp == '\n') { if (bad == 0) { if (++cp < end) continue; else break; } } else break; if (*cp == '\0' || ++cp >= end) *list = NULL; else *list = cp; return (trunc_page(bad)); } printf("Garbage in RAM blacklist, skipping\n"); *list = NULL; return (0); } /* * vm_page_blacklist_check: * * Iterate through the provided string of blacklist addresses, pulling * each entry out of the physical allocator free list and putting it * onto a list for reporting via the vm.page_blacklist sysctl. */ static void vm_page_blacklist_check(char *list, char *end) { vm_paddr_t pa; vm_page_t m; char *next; int ret; next = list; while (next != NULL) { if ((pa = vm_page_blacklist_next(&next, end)) == 0) continue; m = vm_phys_paddr_to_vm_page(pa); if (m == NULL) continue; mtx_lock(&vm_page_queue_free_mtx); ret = vm_phys_unfree_page(m); mtx_unlock(&vm_page_queue_free_mtx); if (ret == TRUE) { TAILQ_INSERT_TAIL(&blacklist_head, m, listq); if (bootverbose) printf("Skipping page with pa 0x%jx\n", (uintmax_t)pa); } } } /* * vm_page_blacklist_load: * * Search for a special module named "ram_blacklist". It'll be a * plain text file provided by the user via the loader directive * of the same name. */ static void vm_page_blacklist_load(char **list, char **end) { void *mod; u_char *ptr; u_int len; mod = NULL; ptr = NULL; mod = preload_search_by_type("ram_blacklist"); if (mod != NULL) { ptr = preload_fetch_addr(mod); len = preload_fetch_size(mod); } *list = ptr; if (ptr != NULL) *end = ptr + len; else *end = NULL; return; } static int sysctl_vm_page_blacklist(SYSCTL_HANDLER_ARGS) { vm_page_t m; struct sbuf sbuf; int error, first; first = 1; error = sysctl_wire_old_buffer(req, 0); if (error != 0) return (error); sbuf_new_for_sysctl(&sbuf, NULL, 128, req); TAILQ_FOREACH(m, &blacklist_head, listq) { sbuf_printf(&sbuf, "%s%#jx", first ? "" : ",", (uintmax_t)m->phys_addr); first = 0; } error = sbuf_finish(&sbuf); sbuf_delete(&sbuf); return (error); } static void vm_page_domain_init(struct vm_domain *vmd) { struct vm_pagequeue *pq; int i; *__DECONST(char **, &vmd->vmd_pagequeues[PQ_INACTIVE].pq_name) = "vm inactive pagequeue"; *__DECONST(int **, &vmd->vmd_pagequeues[PQ_INACTIVE].pq_vcnt) = &vm_cnt.v_inactive_count; *__DECONST(char **, &vmd->vmd_pagequeues[PQ_ACTIVE].pq_name) = "vm active pagequeue"; *__DECONST(int **, &vmd->vmd_pagequeues[PQ_ACTIVE].pq_vcnt) = &vm_cnt.v_active_count; vmd->vmd_page_count = 0; vmd->vmd_free_count = 0; vmd->vmd_segs = 0; vmd->vmd_oom = FALSE; vmd->vmd_pass = 0; for (i = 0; i < PQ_COUNT; i++) { pq = &vmd->vmd_pagequeues[i]; TAILQ_INIT(&pq->pq_pl); mtx_init(&pq->pq_mutex, pq->pq_name, "vm pagequeue", MTX_DEF | MTX_DUPOK); } } /* * vm_page_startup: * * Initializes the resident memory module. * * Allocates memory for the page cells, and * for the object/offset-to-page hash table headers. * Each page cell is initialized and placed on the free list. */ vm_offset_t vm_page_startup(vm_offset_t vaddr) { vm_offset_t mapped; vm_paddr_t page_range; vm_paddr_t new_end; int i; vm_paddr_t pa; vm_paddr_t last_pa; char *list, *listend; vm_paddr_t end; vm_paddr_t biggestsize; vm_paddr_t low_water, high_water; int biggestone; biggestsize = 0; biggestone = 0; vaddr = round_page(vaddr); for (i = 0; phys_avail[i + 1]; i += 2) { phys_avail[i] = round_page(phys_avail[i]); phys_avail[i + 1] = trunc_page(phys_avail[i + 1]); } low_water = phys_avail[0]; high_water = phys_avail[1]; for (i = 0; i < vm_phys_nsegs; i++) { if (vm_phys_segs[i].start < low_water) low_water = vm_phys_segs[i].start; if (vm_phys_segs[i].end > high_water) high_water = vm_phys_segs[i].end; } for (i = 0; phys_avail[i + 1]; i += 2) { vm_paddr_t size = phys_avail[i + 1] - phys_avail[i]; if (size > biggestsize) { biggestone = i; biggestsize = size; } if (phys_avail[i] < low_water) low_water = phys_avail[i]; if (phys_avail[i + 1] > high_water) high_water = phys_avail[i + 1]; } end = phys_avail[biggestone+1]; /* * Initialize the page and queue locks. */ mtx_init(&vm_page_queue_free_mtx, "vm page free queue", NULL, MTX_DEF); for (i = 0; i < PA_LOCK_COUNT; i++) mtx_init(&pa_lock[i], "vm page", NULL, MTX_DEF); for (i = 0; i < vm_ndomains; i++) vm_page_domain_init(&vm_dom[i]); /* * Allocate memory for use when boot strapping the kernel memory * allocator. * * CTFLAG_RDTUN doesn't work during the early boot process, so we must * manually fetch the value. */ TUNABLE_INT_FETCH("vm.boot_pages", &boot_pages); new_end = end - (boot_pages * UMA_SLAB_SIZE); new_end = trunc_page(new_end); mapped = pmap_map(&vaddr, new_end, end, VM_PROT_READ | VM_PROT_WRITE); bzero((void *)mapped, end - new_end); uma_startup((void *)mapped, boot_pages); #if defined(__amd64__) || defined(__i386__) || defined(__arm__) || \ defined(__mips__) /* * Allocate a bitmap to indicate that a random physical page * needs to be included in a minidump. * * The amd64 port needs this to indicate which direct map pages * need to be dumped, via calls to dump_add_page()/dump_drop_page(). * * However, i386 still needs this workspace internally within the * minidump code. In theory, they are not needed on i386, but are * included should the sf_buf code decide to use them. */ last_pa = 0; for (i = 0; dump_avail[i + 1] != 0; i += 2) if (dump_avail[i + 1] > last_pa) last_pa = dump_avail[i + 1]; page_range = last_pa / PAGE_SIZE; vm_page_dump_size = round_page(roundup2(page_range, NBBY) / NBBY); new_end -= vm_page_dump_size; vm_page_dump = (void *)(uintptr_t)pmap_map(&vaddr, new_end, new_end + vm_page_dump_size, VM_PROT_READ | VM_PROT_WRITE); bzero((void *)vm_page_dump, vm_page_dump_size); #endif #ifdef __amd64__ /* * Request that the physical pages underlying the message buffer be * included in a crash dump. Since the message buffer is accessed * through the direct map, they are not automatically included. */ pa = DMAP_TO_PHYS((vm_offset_t)msgbufp->msg_ptr); last_pa = pa + round_page(msgbufsize); while (pa < last_pa) { dump_add_page(pa); pa += PAGE_SIZE; } #endif /* * Compute the number of pages of memory that will be available for * use (taking into account the overhead of a page structure per * page). */ first_page = low_water / PAGE_SIZE; #ifdef VM_PHYSSEG_SPARSE page_range = 0; for (i = 0; i < vm_phys_nsegs; i++) { page_range += atop(vm_phys_segs[i].end - vm_phys_segs[i].start); } for (i = 0; phys_avail[i + 1] != 0; i += 2) page_range += atop(phys_avail[i + 1] - phys_avail[i]); #elif defined(VM_PHYSSEG_DENSE) page_range = high_water / PAGE_SIZE - first_page; #else #error "Either VM_PHYSSEG_DENSE or VM_PHYSSEG_SPARSE must be defined." #endif end = new_end; /* * Reserve an unmapped guard page to trap access to vm_page_array[-1]. */ vaddr += PAGE_SIZE; /* * Initialize the mem entry structures now, and put them in the free * queue. */ new_end = trunc_page(end - page_range * sizeof(struct vm_page)); mapped = pmap_map(&vaddr, new_end, end, VM_PROT_READ | VM_PROT_WRITE); vm_page_array = (vm_page_t) mapped; #if VM_NRESERVLEVEL > 0 /* * Allocate memory for the reservation management system's data * structures. */ new_end = vm_reserv_startup(&vaddr, new_end, high_water); #endif #if defined(__amd64__) || defined(__mips__) /* * pmap_map on amd64 and mips can come out of the direct-map, not kvm * like i386, so the pages must be tracked for a crashdump to include * this data. This includes the vm_page_array and the early UMA * bootstrap pages. */ for (pa = new_end; pa < phys_avail[biggestone + 1]; pa += PAGE_SIZE) dump_add_page(pa); #endif phys_avail[biggestone + 1] = new_end; /* * Add physical memory segments corresponding to the available * physical pages. */ for (i = 0; phys_avail[i + 1] != 0; i += 2) vm_phys_add_seg(phys_avail[i], phys_avail[i + 1]); /* * Clear all of the page structures */ bzero((caddr_t) vm_page_array, page_range * sizeof(struct vm_page)); for (i = 0; i < page_range; i++) vm_page_array[i].order = VM_NFREEORDER; vm_page_array_size = page_range; /* * Initialize the physical memory allocator. */ vm_phys_init(); /* * Add every available physical page that is not blacklisted to * the free lists. */ vm_cnt.v_page_count = 0; vm_cnt.v_free_count = 0; for (i = 0; phys_avail[i + 1] != 0; i += 2) { pa = phys_avail[i]; last_pa = phys_avail[i + 1]; while (pa < last_pa) { vm_phys_add_page(pa); pa += PAGE_SIZE; } } TAILQ_INIT(&blacklist_head); vm_page_blacklist_load(&list, &listend); vm_page_blacklist_check(list, listend); list = kern_getenv("vm.blacklist"); vm_page_blacklist_check(list, NULL); freeenv(list); #if VM_NRESERVLEVEL > 0 /* * Initialize the reservation management system. */ vm_reserv_init(); #endif return (vaddr); } void vm_page_reference(vm_page_t m) { vm_page_aflag_set(m, PGA_REFERENCED); } /* * vm_page_busy_downgrade: * * Downgrade an exclusive busy page into a single shared busy page. */ void vm_page_busy_downgrade(vm_page_t m) { u_int x; vm_page_assert_xbusied(m); for (;;) { x = m->busy_lock; x &= VPB_BIT_WAITERS; if (atomic_cmpset_rel_int(&m->busy_lock, VPB_SINGLE_EXCLUSIVER | x, VPB_SHARERS_WORD(1) | x)) break; } } /* * vm_page_sbusied: * * Return a positive value if the page is shared busied, 0 otherwise. */ int vm_page_sbusied(vm_page_t m) { u_int x; x = m->busy_lock; return ((x & VPB_BIT_SHARED) != 0 && x != VPB_UNBUSIED); } /* * vm_page_sunbusy: * * Shared unbusy a page. */ void vm_page_sunbusy(vm_page_t m) { u_int x; vm_page_assert_sbusied(m); for (;;) { x = m->busy_lock; if (VPB_SHARERS(x) > 1) { if (atomic_cmpset_int(&m->busy_lock, x, x - VPB_ONE_SHARER)) break; continue; } if ((x & VPB_BIT_WAITERS) == 0) { KASSERT(x == VPB_SHARERS_WORD(1), ("vm_page_sunbusy: invalid lock state")); if (atomic_cmpset_int(&m->busy_lock, VPB_SHARERS_WORD(1), VPB_UNBUSIED)) break; continue; } KASSERT(x == (VPB_SHARERS_WORD(1) | VPB_BIT_WAITERS), ("vm_page_sunbusy: invalid lock state for waiters")); vm_page_lock(m); if (!atomic_cmpset_int(&m->busy_lock, x, VPB_UNBUSIED)) { vm_page_unlock(m); continue; } wakeup(m); vm_page_unlock(m); break; } } /* * vm_page_busy_sleep: * * Sleep and release the page lock, using the page pointer as wchan. * This is used to implement the hard-path of busying mechanism. * * The given page must be locked. */ void vm_page_busy_sleep(vm_page_t m, const char *wmesg) { u_int x; vm_page_lock_assert(m, MA_OWNED); x = m->busy_lock; if (x == VPB_UNBUSIED) { vm_page_unlock(m); return; } if ((x & VPB_BIT_WAITERS) == 0 && !atomic_cmpset_int(&m->busy_lock, x, x | VPB_BIT_WAITERS)) { vm_page_unlock(m); return; } msleep(m, vm_page_lockptr(m), PVM | PDROP, wmesg, 0); } /* * vm_page_trysbusy: * * Try to shared busy a page. * If the operation succeeds 1 is returned otherwise 0. * The operation never sleeps. */ int vm_page_trysbusy(vm_page_t m) { u_int x; for (;;) { x = m->busy_lock; if ((x & VPB_BIT_SHARED) == 0) return (0); if (atomic_cmpset_acq_int(&m->busy_lock, x, x + VPB_ONE_SHARER)) return (1); } } /* * vm_page_xunbusy_hard: * * Called after the first try the exclusive unbusy of a page failed. * It is assumed that the waiters bit is on. */ void vm_page_xunbusy_hard(vm_page_t m) { vm_page_assert_xbusied(m); vm_page_lock(m); atomic_store_rel_int(&m->busy_lock, VPB_UNBUSIED); wakeup(m); vm_page_unlock(m); } /* * vm_page_flash: * * Wakeup anyone waiting for the page. * The ownership bits do not change. * * The given page must be locked. */ void vm_page_flash(vm_page_t m) { u_int x; vm_page_lock_assert(m, MA_OWNED); for (;;) { x = m->busy_lock; if ((x & VPB_BIT_WAITERS) == 0) return; if (atomic_cmpset_int(&m->busy_lock, x, x & (~VPB_BIT_WAITERS))) break; } wakeup(m); } /* * Keep page from being freed by the page daemon * much of the same effect as wiring, except much lower * overhead and should be used only for *very* temporary * holding ("wiring"). */ void vm_page_hold(vm_page_t mem) { vm_page_lock_assert(mem, MA_OWNED); mem->hold_count++; } void vm_page_unhold(vm_page_t mem) { vm_page_lock_assert(mem, MA_OWNED); KASSERT(mem->hold_count >= 1, ("vm_page_unhold: hold count < 0!!!")); --mem->hold_count; if (mem->hold_count == 0 && (mem->flags & PG_UNHOLDFREE) != 0) vm_page_free_toq(mem); } /* * vm_page_unhold_pages: * * Unhold each of the pages that is referenced by the given array. */ void vm_page_unhold_pages(vm_page_t *ma, int count) { struct mtx *mtx, *new_mtx; mtx = NULL; for (; count != 0; count--) { /* * Avoid releasing and reacquiring the same page lock. */ new_mtx = vm_page_lockptr(*ma); if (mtx != new_mtx) { if (mtx != NULL) mtx_unlock(mtx); mtx = new_mtx; mtx_lock(mtx); } vm_page_unhold(*ma); ma++; } if (mtx != NULL) mtx_unlock(mtx); } vm_page_t PHYS_TO_VM_PAGE(vm_paddr_t pa) { vm_page_t m; #ifdef VM_PHYSSEG_SPARSE m = vm_phys_paddr_to_vm_page(pa); if (m == NULL) m = vm_phys_fictitious_to_vm_page(pa); return (m); #elif defined(VM_PHYSSEG_DENSE) long pi; pi = atop(pa); if (pi >= first_page && (pi - first_page) < vm_page_array_size) { m = &vm_page_array[pi - first_page]; return (m); } return (vm_phys_fictitious_to_vm_page(pa)); #else #error "Either VM_PHYSSEG_DENSE or VM_PHYSSEG_SPARSE must be defined." #endif } /* * vm_page_getfake: * * Create a fictitious page with the specified physical address and * memory attribute. The memory attribute is the only the machine- * dependent aspect of a fictitious page that must be initialized. */ vm_page_t vm_page_getfake(vm_paddr_t paddr, vm_memattr_t memattr) { vm_page_t m; m = uma_zalloc(fakepg_zone, M_WAITOK | M_ZERO); vm_page_initfake(m, paddr, memattr); return (m); } void vm_page_initfake(vm_page_t m, vm_paddr_t paddr, vm_memattr_t memattr) { if ((m->flags & PG_FICTITIOUS) != 0) { /* * The page's memattr might have changed since the * previous initialization. Update the pmap to the * new memattr. */ goto memattr; } m->phys_addr = paddr; m->queue = PQ_NONE; /* Fictitious pages don't use "segind". */ m->flags = PG_FICTITIOUS; /* Fictitious pages don't use "order" or "pool". */ m->oflags = VPO_UNMANAGED; m->busy_lock = VPB_SINGLE_EXCLUSIVER; m->wire_count = 1; pmap_page_init(m); memattr: pmap_page_set_memattr(m, memattr); } /* * vm_page_putfake: * * Release a fictitious page. */ void vm_page_putfake(vm_page_t m) { KASSERT((m->oflags & VPO_UNMANAGED) != 0, ("managed %p", m)); KASSERT((m->flags & PG_FICTITIOUS) != 0, ("vm_page_putfake: bad page %p", m)); uma_zfree(fakepg_zone, m); } /* * vm_page_updatefake: * * Update the given fictitious page to the specified physical address and * memory attribute. */ void vm_page_updatefake(vm_page_t m, vm_paddr_t paddr, vm_memattr_t memattr) { KASSERT((m->flags & PG_FICTITIOUS) != 0, ("vm_page_updatefake: bad page %p", m)); m->phys_addr = paddr; pmap_page_set_memattr(m, memattr); } /* * vm_page_free: * * Free a page. */ void vm_page_free(vm_page_t m) { m->flags &= ~PG_ZERO; vm_page_free_toq(m); } /* * vm_page_free_zero: * * Free a page to the zerod-pages queue */ void vm_page_free_zero(vm_page_t m) { m->flags |= PG_ZERO; vm_page_free_toq(m); } /* * Unbusy and handle the page queueing for a page from the VOP_GETPAGES() * array which is not the request page. */ void vm_page_readahead_finish(vm_page_t m) { if (m->valid != 0) { /* * Since the page is not the requested page, whether * it should be activated or deactivated is not * obvious. Empirical results have shown that * deactivating the page is usually the best choice, * unless the page is wanted by another thread. */ vm_page_lock(m); if ((m->busy_lock & VPB_BIT_WAITERS) != 0) vm_page_activate(m); else vm_page_deactivate(m); vm_page_unlock(m); vm_page_xunbusy(m); } else { /* * Free the completely invalid page. Such page state * occurs due to the short read operation which did * not covered our page at all, or in case when a read * error happens. */ vm_page_lock(m); vm_page_free(m); vm_page_unlock(m); } } /* * vm_page_sleep_if_busy: * * Sleep and release the page queues lock if the page is busied. * Returns TRUE if the thread slept. * * The given page must be unlocked and object containing it must * be locked. */ int vm_page_sleep_if_busy(vm_page_t m, const char *msg) { vm_object_t obj; vm_page_lock_assert(m, MA_NOTOWNED); VM_OBJECT_ASSERT_WLOCKED(m->object); if (vm_page_busied(m)) { /* * The page-specific object must be cached because page * identity can change during the sleep, causing the * re-lock of a different object. * It is assumed that a reference to the object is already * held by the callers. */ obj = m->object; vm_page_lock(m); VM_OBJECT_WUNLOCK(obj); vm_page_busy_sleep(m, msg); VM_OBJECT_WLOCK(obj); return (TRUE); } return (FALSE); } /* * vm_page_dirty_KBI: [ internal use only ] * * Set all bits in the page's dirty field. * * The object containing the specified page must be locked if the * call is made from the machine-independent layer. * * See vm_page_clear_dirty_mask(). * * This function should only be called by vm_page_dirty(). */ void vm_page_dirty_KBI(vm_page_t m) { /* These assertions refer to this operation by its public name. */ KASSERT((m->flags & PG_CACHED) == 0, ("vm_page_dirty: page in cache!")); KASSERT(m->valid == VM_PAGE_BITS_ALL, ("vm_page_dirty: page is invalid!")); m->dirty = VM_PAGE_BITS_ALL; } /* * vm_page_insert: [ internal use only ] * * Inserts the given mem entry into the object and object list. * * The object must be locked. */ int vm_page_insert(vm_page_t m, vm_object_t object, vm_pindex_t pindex) { vm_page_t mpred; VM_OBJECT_ASSERT_WLOCKED(object); mpred = vm_radix_lookup_le(&object->rtree, pindex); return (vm_page_insert_after(m, object, pindex, mpred)); } /* * vm_page_insert_after: * * Inserts the page "m" into the specified object at offset "pindex". * * The page "mpred" must immediately precede the offset "pindex" within * the specified object. * * The object must be locked. */ static int vm_page_insert_after(vm_page_t m, vm_object_t object, vm_pindex_t pindex, vm_page_t mpred) { vm_pindex_t sidx; vm_object_t sobj; vm_page_t msucc; VM_OBJECT_ASSERT_WLOCKED(object); KASSERT(m->object == NULL, ("vm_page_insert_after: page already inserted")); if (mpred != NULL) { KASSERT(mpred->object == object, ("vm_page_insert_after: object doesn't contain mpred")); KASSERT(mpred->pindex < pindex, ("vm_page_insert_after: mpred doesn't precede pindex")); msucc = TAILQ_NEXT(mpred, listq); } else msucc = TAILQ_FIRST(&object->memq); if (msucc != NULL) KASSERT(msucc->pindex > pindex, ("vm_page_insert_after: msucc doesn't succeed pindex")); /* * Record the object/offset pair in this page */ sobj = m->object; sidx = m->pindex; m->object = object; m->pindex = pindex; /* * Now link into the object's ordered list of backed pages. */ if (vm_radix_insert(&object->rtree, m)) { m->object = sobj; m->pindex = sidx; return (1); } vm_page_insert_radixdone(m, object, mpred); return (0); } /* * vm_page_insert_radixdone: * * Complete page "m" insertion into the specified object after the * radix trie hooking. * * The page "mpred" must precede the offset "m->pindex" within the * specified object. * * The object must be locked. */ static void vm_page_insert_radixdone(vm_page_t m, vm_object_t object, vm_page_t mpred) { VM_OBJECT_ASSERT_WLOCKED(object); KASSERT(object != NULL && m->object == object, ("vm_page_insert_radixdone: page %p has inconsistent object", m)); if (mpred != NULL) { KASSERT(mpred->object == object, ("vm_page_insert_after: object doesn't contain mpred")); KASSERT(mpred->pindex < m->pindex, ("vm_page_insert_after: mpred doesn't precede pindex")); } if (mpred != NULL) TAILQ_INSERT_AFTER(&object->memq, mpred, m, listq); else TAILQ_INSERT_HEAD(&object->memq, m, listq); /* * Show that the object has one more resident page. */ object->resident_page_count++; /* * Hold the vnode until the last page is released. */ if (object->resident_page_count == 1 && object->type == OBJT_VNODE) vhold(object->handle); /* * Since we are inserting a new and possibly dirty page, * update the object's OBJ_MIGHTBEDIRTY flag. */ if (pmap_page_is_write_mapped(m)) vm_object_set_writeable_dirty(object); } /* * vm_page_remove: * * Removes the given mem entry from the object/offset-page * table and the object page list, but do not invalidate/terminate * the backing store. * * The object must be locked. The page must be locked if it is managed. */ void vm_page_remove(vm_page_t m) { vm_object_t object; boolean_t lockacq; if ((m->oflags & VPO_UNMANAGED) == 0) vm_page_lock_assert(m, MA_OWNED); if ((object = m->object) == NULL) return; VM_OBJECT_ASSERT_WLOCKED(object); if (vm_page_xbusied(m)) { lockacq = FALSE; if ((m->oflags & VPO_UNMANAGED) != 0 && !mtx_owned(vm_page_lockptr(m))) { lockacq = TRUE; vm_page_lock(m); } vm_page_flash(m); atomic_store_rel_int(&m->busy_lock, VPB_UNBUSIED); if (lockacq) vm_page_unlock(m); } /* * Now remove from the object's list of backed pages. */ vm_radix_remove(&object->rtree, m->pindex); TAILQ_REMOVE(&object->memq, m, listq); /* * And show that the object has one fewer resident page. */ object->resident_page_count--; /* * The vnode may now be recycled. */ if (object->resident_page_count == 0 && object->type == OBJT_VNODE) vdrop(object->handle); m->object = NULL; } /* * vm_page_lookup: * * Returns the page associated with the object/offset * pair specified; if none is found, NULL is returned. * * The object must be locked. */ vm_page_t vm_page_lookup(vm_object_t object, vm_pindex_t pindex) { VM_OBJECT_ASSERT_LOCKED(object); return (vm_radix_lookup(&object->rtree, pindex)); } /* * vm_page_find_least: * * Returns the page associated with the object with least pindex * greater than or equal to the parameter pindex, or NULL. * * The object must be locked. */ vm_page_t vm_page_find_least(vm_object_t object, vm_pindex_t pindex) { vm_page_t m; VM_OBJECT_ASSERT_LOCKED(object); if ((m = TAILQ_FIRST(&object->memq)) != NULL && m->pindex < pindex) m = vm_radix_lookup_ge(&object->rtree, pindex); return (m); } /* * Returns the given page's successor (by pindex) within the object if it is * resident; if none is found, NULL is returned. * * The object must be locked. */ vm_page_t vm_page_next(vm_page_t m) { vm_page_t next; VM_OBJECT_ASSERT_WLOCKED(m->object); if ((next = TAILQ_NEXT(m, listq)) != NULL && next->pindex != m->pindex + 1) next = NULL; return (next); } /* * Returns the given page's predecessor (by pindex) within the object if it is * resident; if none is found, NULL is returned. * * The object must be locked. */ vm_page_t vm_page_prev(vm_page_t m) { vm_page_t prev; VM_OBJECT_ASSERT_WLOCKED(m->object); if ((prev = TAILQ_PREV(m, pglist, listq)) != NULL && prev->pindex != m->pindex - 1) prev = NULL; return (prev); } /* * Uses the page mnew as a replacement for an existing page at index * pindex which must be already present in the object. * * The existing page must not be on a paging queue. */ vm_page_t vm_page_replace(vm_page_t mnew, vm_object_t object, vm_pindex_t pindex) { vm_page_t mold, mpred; VM_OBJECT_ASSERT_WLOCKED(object); /* * This function mostly follows vm_page_insert() and * vm_page_remove() without the radix, object count and vnode * dance. Double check such functions for more comments. */ mpred = vm_radix_lookup(&object->rtree, pindex); KASSERT(mpred != NULL, ("vm_page_replace: replacing page not present with pindex")); mpred = TAILQ_PREV(mpred, respgs, listq); if (mpred != NULL) KASSERT(mpred->pindex < pindex, ("vm_page_insert_after: mpred doesn't precede pindex")); mnew->object = object; mnew->pindex = pindex; mold = vm_radix_replace(&object->rtree, mnew); KASSERT(mold->queue == PQ_NONE, ("vm_page_replace: mold is on a paging queue")); /* Detach the old page from the resident tailq. */ TAILQ_REMOVE(&object->memq, mold, listq); mold->object = NULL; vm_page_xunbusy(mold); /* Insert the new page in the resident tailq. */ if (mpred != NULL) TAILQ_INSERT_AFTER(&object->memq, mpred, mnew, listq); else TAILQ_INSERT_HEAD(&object->memq, mnew, listq); if (pmap_page_is_write_mapped(mnew)) vm_object_set_writeable_dirty(object); return (mold); } /* * vm_page_rename: * * Move the given memory entry from its * current object to the specified target object/offset. * * Note: swap associated with the page must be invalidated by the move. We * have to do this for several reasons: (1) we aren't freeing the * page, (2) we are dirtying the page, (3) the VM system is probably * moving the page from object A to B, and will then later move * the backing store from A to B and we can't have a conflict. * * Note: we *always* dirty the page. It is necessary both for the * fact that we moved it, and because we may be invalidating * swap. If the page is on the cache, we have to deactivate it * or vm_page_dirty() will panic. Dirty pages are not allowed * on the cache. * * The objects must be locked. */ int vm_page_rename(vm_page_t m, vm_object_t new_object, vm_pindex_t new_pindex) { vm_page_t mpred; vm_pindex_t opidx; VM_OBJECT_ASSERT_WLOCKED(new_object); mpred = vm_radix_lookup_le(&new_object->rtree, new_pindex); KASSERT(mpred == NULL || mpred->pindex != new_pindex, ("vm_page_rename: pindex already renamed")); /* * Create a custom version of vm_page_insert() which does not depend * by m_prev and can cheat on the implementation aspects of the * function. */ opidx = m->pindex; m->pindex = new_pindex; if (vm_radix_insert(&new_object->rtree, m)) { m->pindex = opidx; return (1); } /* * The operation cannot fail anymore. The removal must happen before * the listq iterator is tainted. */ m->pindex = opidx; vm_page_lock(m); vm_page_remove(m); /* Return back to the new pindex to complete vm_page_insert(). */ m->pindex = new_pindex; m->object = new_object; vm_page_unlock(m); vm_page_insert_radixdone(m, new_object, mpred); vm_page_dirty(m); return (0); } /* * Convert all of the given object's cached pages that have a * pindex within the given range into free pages. If the value * zero is given for "end", then the range's upper bound is * infinity. If the given object is backed by a vnode and it * transitions from having one or more cached pages to none, the * vnode's hold count is reduced. */ void vm_page_cache_free(vm_object_t object, vm_pindex_t start, vm_pindex_t end) { vm_page_t m; boolean_t empty; mtx_lock(&vm_page_queue_free_mtx); if (__predict_false(vm_radix_is_empty(&object->cache))) { mtx_unlock(&vm_page_queue_free_mtx); return; } while ((m = vm_radix_lookup_ge(&object->cache, start)) != NULL) { if (end != 0 && m->pindex >= end) break; vm_radix_remove(&object->cache, m->pindex); vm_page_cache_turn_free(m); } empty = vm_radix_is_empty(&object->cache); mtx_unlock(&vm_page_queue_free_mtx); if (object->type == OBJT_VNODE && empty) vdrop(object->handle); } /* * Returns the cached page that is associated with the given * object and offset. If, however, none exists, returns NULL. * * The free page queue must be locked. */ static inline vm_page_t vm_page_cache_lookup(vm_object_t object, vm_pindex_t pindex) { mtx_assert(&vm_page_queue_free_mtx, MA_OWNED); return (vm_radix_lookup(&object->cache, pindex)); } /* * Remove the given cached page from its containing object's * collection of cached pages. * * The free page queue must be locked. */ static void vm_page_cache_remove(vm_page_t m) { mtx_assert(&vm_page_queue_free_mtx, MA_OWNED); KASSERT((m->flags & PG_CACHED) != 0, ("vm_page_cache_remove: page %p is not cached", m)); vm_radix_remove(&m->object->cache, m->pindex); m->object = NULL; vm_cnt.v_cache_count--; } /* * Transfer all of the cached pages with offset greater than or * equal to 'offidxstart' from the original object's cache to the * new object's cache. However, any cached pages with offset * greater than or equal to the new object's size are kept in the * original object. Initially, the new object's cache must be * empty. Offset 'offidxstart' in the original object must * correspond to offset zero in the new object. * * The new object must be locked. */ void vm_page_cache_transfer(vm_object_t orig_object, vm_pindex_t offidxstart, vm_object_t new_object) { vm_page_t m; /* * Insertion into an object's collection of cached pages * requires the object to be locked. In contrast, removal does * not. */ VM_OBJECT_ASSERT_WLOCKED(new_object); KASSERT(vm_radix_is_empty(&new_object->cache), ("vm_page_cache_transfer: object %p has cached pages", new_object)); mtx_lock(&vm_page_queue_free_mtx); while ((m = vm_radix_lookup_ge(&orig_object->cache, offidxstart)) != NULL) { /* * Transfer all of the pages with offset greater than or * equal to 'offidxstart' from the original object's * cache to the new object's cache. */ if ((m->pindex - offidxstart) >= new_object->size) break; vm_radix_remove(&orig_object->cache, m->pindex); /* Update the page's object and offset. */ m->object = new_object; m->pindex -= offidxstart; if (vm_radix_insert(&new_object->cache, m)) vm_page_cache_turn_free(m); } mtx_unlock(&vm_page_queue_free_mtx); } /* * Returns TRUE if a cached page is associated with the given object and * offset, and FALSE otherwise. * * The object must be locked. */ boolean_t vm_page_is_cached(vm_object_t object, vm_pindex_t pindex) { vm_page_t m; /* * Insertion into an object's collection of cached pages requires the * object to be locked. Therefore, if the object is locked and the * object's collection is empty, there is no need to acquire the free * page queues lock in order to prove that the specified page doesn't * exist. */ VM_OBJECT_ASSERT_WLOCKED(object); if (__predict_true(vm_object_cache_is_empty(object))) return (FALSE); mtx_lock(&vm_page_queue_free_mtx); m = vm_page_cache_lookup(object, pindex); mtx_unlock(&vm_page_queue_free_mtx); return (m != NULL); } /* * vm_page_alloc: * * Allocate and return a page that is associated with the specified * object and offset pair. By default, this page is exclusive busied. * * The caller must always specify an allocation class. * * allocation classes: * VM_ALLOC_NORMAL normal process request * VM_ALLOC_SYSTEM system *really* needs a page * VM_ALLOC_INTERRUPT interrupt time request * * optional allocation flags: * VM_ALLOC_COUNT(number) the number of additional pages that the caller * intends to allocate * VM_ALLOC_IFCACHED return page only if it is cached * VM_ALLOC_IFNOTCACHED return NULL, do not reactivate if the page * is cached * VM_ALLOC_NOBUSY do not exclusive busy the page * VM_ALLOC_NODUMP do not include the page in a kernel core dump * VM_ALLOC_NOOBJ page is not associated with an object and * should not be exclusive busy * VM_ALLOC_SBUSY shared busy the allocated page * VM_ALLOC_WIRED wire the allocated page * VM_ALLOC_ZERO prefer a zeroed page * * This routine may not sleep. */ vm_page_t vm_page_alloc(vm_object_t object, vm_pindex_t pindex, int req) { struct vnode *vp = NULL; vm_object_t m_object; vm_page_t m, mpred; int flags, req_class; mpred = 0; /* XXX: pacify gcc */ KASSERT((object != NULL) == ((req & VM_ALLOC_NOOBJ) == 0) && (object != NULL || (req & VM_ALLOC_SBUSY) == 0) && ((req & (VM_ALLOC_NOBUSY | VM_ALLOC_SBUSY)) != (VM_ALLOC_NOBUSY | VM_ALLOC_SBUSY)), ("vm_page_alloc: inconsistent object(%p)/req(%x)", (void *)object, req)); if (object != NULL) VM_OBJECT_ASSERT_WLOCKED(object); req_class = req & VM_ALLOC_CLASS_MASK; /* * The page daemon is allowed to dig deeper into the free page list. */ if (curproc == pageproc && req_class != VM_ALLOC_INTERRUPT) req_class = VM_ALLOC_SYSTEM; if (object != NULL) { mpred = vm_radix_lookup_le(&object->rtree, pindex); KASSERT(mpred == NULL || mpred->pindex != pindex, ("vm_page_alloc: pindex already allocated")); } /* * The page allocation request can came from consumers which already * hold the free page queue mutex, like vm_page_insert() in * vm_page_cache(). */ mtx_lock_flags(&vm_page_queue_free_mtx, MTX_RECURSE); if (vm_cnt.v_free_count + vm_cnt.v_cache_count > vm_cnt.v_free_reserved || (req_class == VM_ALLOC_SYSTEM && vm_cnt.v_free_count + vm_cnt.v_cache_count > vm_cnt.v_interrupt_free_min) || (req_class == VM_ALLOC_INTERRUPT && vm_cnt.v_free_count + vm_cnt.v_cache_count > 0)) { /* * Allocate from the free queue if the number of free pages * exceeds the minimum for the request class. */ if (object != NULL && (m = vm_page_cache_lookup(object, pindex)) != NULL) { if ((req & VM_ALLOC_IFNOTCACHED) != 0) { mtx_unlock(&vm_page_queue_free_mtx); return (NULL); } if (vm_phys_unfree_page(m)) vm_phys_set_pool(VM_FREEPOOL_DEFAULT, m, 0); #if VM_NRESERVLEVEL > 0 else if (!vm_reserv_reactivate_page(m)) #else else #endif panic("vm_page_alloc: cache page %p is missing" " from the free queue", m); } else if ((req & VM_ALLOC_IFCACHED) != 0) { mtx_unlock(&vm_page_queue_free_mtx); return (NULL); #if VM_NRESERVLEVEL > 0 } else if (object == NULL || (object->flags & (OBJ_COLORED | OBJ_FICTITIOUS)) != OBJ_COLORED || (m = vm_reserv_alloc_page(object, pindex, mpred)) == NULL) { #else } else { #endif m = vm_phys_alloc_pages(object != NULL ? VM_FREEPOOL_DEFAULT : VM_FREEPOOL_DIRECT, 0); #if VM_NRESERVLEVEL > 0 if (m == NULL && vm_reserv_reclaim_inactive()) { m = vm_phys_alloc_pages(object != NULL ? VM_FREEPOOL_DEFAULT : VM_FREEPOOL_DIRECT, 0); } #endif } } else { /* * Not allocatable, give up. */ mtx_unlock(&vm_page_queue_free_mtx); atomic_add_int(&vm_pageout_deficit, max((u_int)req >> VM_ALLOC_COUNT_SHIFT, 1)); pagedaemon_wakeup(); return (NULL); } /* * At this point we had better have found a good page. */ KASSERT(m != NULL, ("vm_page_alloc: missing page")); KASSERT(m->queue == PQ_NONE, ("vm_page_alloc: page %p has unexpected queue %d", m, m->queue)); KASSERT(m->wire_count == 0, ("vm_page_alloc: page %p is wired", m)); KASSERT(m->hold_count == 0, ("vm_page_alloc: page %p is held", m)); KASSERT(!vm_page_sbusied(m), ("vm_page_alloc: page %p is busy", m)); KASSERT(m->dirty == 0, ("vm_page_alloc: page %p is dirty", m)); KASSERT(pmap_page_get_memattr(m) == VM_MEMATTR_DEFAULT, ("vm_page_alloc: page %p has unexpected memattr %d", m, pmap_page_get_memattr(m))); if ((m->flags & PG_CACHED) != 0) { KASSERT((m->flags & PG_ZERO) == 0, ("vm_page_alloc: cached page %p is PG_ZERO", m)); KASSERT(m->valid != 0, ("vm_page_alloc: cached page %p is invalid", m)); if (m->object == object && m->pindex == pindex) vm_cnt.v_reactivated++; else m->valid = 0; m_object = m->object; vm_page_cache_remove(m); if (m_object->type == OBJT_VNODE && vm_object_cache_is_empty(m_object)) vp = m_object->handle; } else { KASSERT(m->valid == 0, ("vm_page_alloc: free page %p is valid", m)); vm_phys_freecnt_adj(m, -1); if ((m->flags & PG_ZERO) != 0) vm_page_zero_count--; } mtx_unlock(&vm_page_queue_free_mtx); /* * Initialize the page. Only the PG_ZERO flag is inherited. */ flags = 0; if ((req & VM_ALLOC_ZERO) != 0) flags = PG_ZERO; flags &= m->flags; if ((req & VM_ALLOC_NODUMP) != 0) flags |= PG_NODUMP; m->flags = flags; m->aflags = 0; m->oflags = object == NULL || (object->flags & OBJ_UNMANAGED) != 0 ? VPO_UNMANAGED : 0; m->busy_lock = VPB_UNBUSIED; if ((req & (VM_ALLOC_NOBUSY | VM_ALLOC_NOOBJ | VM_ALLOC_SBUSY)) == 0) m->busy_lock = VPB_SINGLE_EXCLUSIVER; if ((req & VM_ALLOC_SBUSY) != 0) m->busy_lock = VPB_SHARERS_WORD(1); if (req & VM_ALLOC_WIRED) { /* * The page lock is not required for wiring a page until that * page is inserted into the object. */ atomic_add_int(&vm_cnt.v_wire_count, 1); m->wire_count = 1; } m->act_count = 0; if (object != NULL) { if (vm_page_insert_after(m, object, pindex, mpred)) { /* See the comment below about hold count. */ if (vp != NULL) vdrop(vp); pagedaemon_wakeup(); if (req & VM_ALLOC_WIRED) { atomic_subtract_int(&vm_cnt.v_wire_count, 1); m->wire_count = 0; } m->object = NULL; m->oflags = VPO_UNMANAGED; vm_page_free(m); return (NULL); } /* Ignore device objects; the pager sets "memattr" for them. */ if (object->memattr != VM_MEMATTR_DEFAULT && (object->flags & OBJ_FICTITIOUS) == 0) pmap_page_set_memattr(m, object->memattr); } else m->pindex = pindex; /* * The following call to vdrop() must come after the above call * to vm_page_insert() in case both affect the same object and * vnode. Otherwise, the affected vnode's hold count could * temporarily become zero. */ if (vp != NULL) vdrop(vp); /* * Don't wakeup too often - wakeup the pageout daemon when * we would be nearly out of memory. */ if (vm_paging_needed()) pagedaemon_wakeup(); return (m); } static void vm_page_alloc_contig_vdrop(struct spglist *lst) { while (!SLIST_EMPTY(lst)) { vdrop((struct vnode *)SLIST_FIRST(lst)-> plinks.s.pv); SLIST_REMOVE_HEAD(lst, plinks.s.ss); } } /* * vm_page_alloc_contig: * * Allocate a contiguous set of physical pages of the given size "npages" * from the free lists. All of the physical pages must be at or above * the given physical address "low" and below the given physical address * "high". The given value "alignment" determines the alignment of the * first physical page in the set. If the given value "boundary" is * non-zero, then the set of physical pages cannot cross any physical * address boundary that is a multiple of that value. Both "alignment" * and "boundary" must be a power of two. * * If the specified memory attribute, "memattr", is VM_MEMATTR_DEFAULT, * then the memory attribute setting for the physical pages is configured * to the object's memory attribute setting. Otherwise, the memory * attribute setting for the physical pages is configured to "memattr", * overriding the object's memory attribute setting. However, if the * object's memory attribute setting is not VM_MEMATTR_DEFAULT, then the * memory attribute setting for the physical pages cannot be configured * to VM_MEMATTR_DEFAULT. * * The caller must always specify an allocation class. * * allocation classes: * VM_ALLOC_NORMAL normal process request * VM_ALLOC_SYSTEM system *really* needs a page * VM_ALLOC_INTERRUPT interrupt time request * * optional allocation flags: * VM_ALLOC_NOBUSY do not exclusive busy the page + * VM_ALLOC_NODUMP do not include the page in a kernel core dump * VM_ALLOC_NOOBJ page is not associated with an object and * should not be exclusive busy * VM_ALLOC_SBUSY shared busy the allocated page * VM_ALLOC_WIRED wire the allocated page * VM_ALLOC_ZERO prefer a zeroed page * * This routine may not sleep. */ vm_page_t vm_page_alloc_contig(vm_object_t object, vm_pindex_t pindex, int req, u_long npages, vm_paddr_t low, vm_paddr_t high, u_long alignment, vm_paddr_t boundary, vm_memattr_t memattr) { struct vnode *drop; struct spglist deferred_vdrop_list; vm_page_t m, m_tmp, m_ret; u_int flags; int req_class; KASSERT((object != NULL) == ((req & VM_ALLOC_NOOBJ) == 0) && (object != NULL || (req & VM_ALLOC_SBUSY) == 0) && ((req & (VM_ALLOC_NOBUSY | VM_ALLOC_SBUSY)) != (VM_ALLOC_NOBUSY | VM_ALLOC_SBUSY)), ("vm_page_alloc: inconsistent object(%p)/req(%x)", (void *)object, req)); if (object != NULL) { VM_OBJECT_ASSERT_WLOCKED(object); KASSERT(object->type == OBJT_PHYS, ("vm_page_alloc_contig: object %p isn't OBJT_PHYS", object)); } KASSERT(npages > 0, ("vm_page_alloc_contig: npages is zero")); req_class = req & VM_ALLOC_CLASS_MASK; /* * The page daemon is allowed to dig deeper into the free page list. */ if (curproc == pageproc && req_class != VM_ALLOC_INTERRUPT) req_class = VM_ALLOC_SYSTEM; SLIST_INIT(&deferred_vdrop_list); mtx_lock(&vm_page_queue_free_mtx); if (vm_cnt.v_free_count + vm_cnt.v_cache_count >= npages + vm_cnt.v_free_reserved || (req_class == VM_ALLOC_SYSTEM && vm_cnt.v_free_count + vm_cnt.v_cache_count >= npages + vm_cnt.v_interrupt_free_min) || (req_class == VM_ALLOC_INTERRUPT && vm_cnt.v_free_count + vm_cnt.v_cache_count >= npages)) { #if VM_NRESERVLEVEL > 0 retry: if (object == NULL || (object->flags & OBJ_COLORED) == 0 || (m_ret = vm_reserv_alloc_contig(object, pindex, npages, low, high, alignment, boundary)) == NULL) #endif m_ret = vm_phys_alloc_contig(npages, low, high, alignment, boundary); } else { mtx_unlock(&vm_page_queue_free_mtx); atomic_add_int(&vm_pageout_deficit, npages); pagedaemon_wakeup(); return (NULL); } if (m_ret != NULL) for (m = m_ret; m < &m_ret[npages]; m++) { drop = vm_page_alloc_init(m); if (drop != NULL) { /* * Enqueue the vnode for deferred vdrop(). */ m->plinks.s.pv = drop; SLIST_INSERT_HEAD(&deferred_vdrop_list, m, plinks.s.ss); } } else { #if VM_NRESERVLEVEL > 0 if (vm_reserv_reclaim_contig(npages, low, high, alignment, boundary)) goto retry; #endif } mtx_unlock(&vm_page_queue_free_mtx); if (m_ret == NULL) return (NULL); /* * Initialize the pages. Only the PG_ZERO flag is inherited. */ flags = 0; if ((req & VM_ALLOC_ZERO) != 0) flags = PG_ZERO; if ((req & VM_ALLOC_NODUMP) != 0) flags |= PG_NODUMP; if ((req & VM_ALLOC_WIRED) != 0) atomic_add_int(&vm_cnt.v_wire_count, npages); if (object != NULL) { if (object->memattr != VM_MEMATTR_DEFAULT && memattr == VM_MEMATTR_DEFAULT) memattr = object->memattr; } for (m = m_ret; m < &m_ret[npages]; m++) { m->aflags = 0; m->flags = (m->flags | PG_NODUMP) & flags; m->busy_lock = VPB_UNBUSIED; if (object != NULL) { if ((req & (VM_ALLOC_NOBUSY | VM_ALLOC_SBUSY)) == 0) m->busy_lock = VPB_SINGLE_EXCLUSIVER; if ((req & VM_ALLOC_SBUSY) != 0) m->busy_lock = VPB_SHARERS_WORD(1); } if ((req & VM_ALLOC_WIRED) != 0) m->wire_count = 1; /* Unmanaged pages don't use "act_count". */ m->oflags = VPO_UNMANAGED; if (object != NULL) { if (vm_page_insert(m, object, pindex)) { vm_page_alloc_contig_vdrop( &deferred_vdrop_list); if (vm_paging_needed()) pagedaemon_wakeup(); if ((req & VM_ALLOC_WIRED) != 0) atomic_subtract_int(&vm_cnt.v_wire_count, npages); for (m_tmp = m, m = m_ret; m < &m_ret[npages]; m++) { if ((req & VM_ALLOC_WIRED) != 0) m->wire_count = 0; if (m >= m_tmp) m->object = NULL; vm_page_free(m); } return (NULL); } } else m->pindex = pindex; if (memattr != VM_MEMATTR_DEFAULT) pmap_page_set_memattr(m, memattr); pindex++; } vm_page_alloc_contig_vdrop(&deferred_vdrop_list); if (vm_paging_needed()) pagedaemon_wakeup(); return (m_ret); } /* * Initialize a page that has been freshly dequeued from a freelist. * The caller has to drop the vnode returned, if it is not NULL. * * This function may only be used to initialize unmanaged pages. * * To be called with vm_page_queue_free_mtx held. */ static struct vnode * vm_page_alloc_init(vm_page_t m) { struct vnode *drop; vm_object_t m_object; KASSERT(m->queue == PQ_NONE, ("vm_page_alloc_init: page %p has unexpected queue %d", m, m->queue)); KASSERT(m->wire_count == 0, ("vm_page_alloc_init: page %p is wired", m)); KASSERT(m->hold_count == 0, ("vm_page_alloc_init: page %p is held", m)); KASSERT(!vm_page_sbusied(m), ("vm_page_alloc_init: page %p is busy", m)); KASSERT(m->dirty == 0, ("vm_page_alloc_init: page %p is dirty", m)); KASSERT(pmap_page_get_memattr(m) == VM_MEMATTR_DEFAULT, ("vm_page_alloc_init: page %p has unexpected memattr %d", m, pmap_page_get_memattr(m))); mtx_assert(&vm_page_queue_free_mtx, MA_OWNED); drop = NULL; if ((m->flags & PG_CACHED) != 0) { KASSERT((m->flags & PG_ZERO) == 0, ("vm_page_alloc_init: cached page %p is PG_ZERO", m)); m->valid = 0; m_object = m->object; vm_page_cache_remove(m); if (m_object->type == OBJT_VNODE && vm_object_cache_is_empty(m_object)) drop = m_object->handle; } else { KASSERT(m->valid == 0, ("vm_page_alloc_init: free page %p is valid", m)); vm_phys_freecnt_adj(m, -1); if ((m->flags & PG_ZERO) != 0) vm_page_zero_count--; } return (drop); } /* * vm_page_alloc_freelist: * * Allocate a physical page from the specified free page list. * * The caller must always specify an allocation class. * * allocation classes: * VM_ALLOC_NORMAL normal process request * VM_ALLOC_SYSTEM system *really* needs a page * VM_ALLOC_INTERRUPT interrupt time request * * optional allocation flags: * VM_ALLOC_COUNT(number) the number of additional pages that the caller * intends to allocate * VM_ALLOC_WIRED wire the allocated page * VM_ALLOC_ZERO prefer a zeroed page * * This routine may not sleep. */ vm_page_t vm_page_alloc_freelist(int flind, int req) { struct vnode *drop; vm_page_t m; u_int flags; int req_class; req_class = req & VM_ALLOC_CLASS_MASK; /* * The page daemon is allowed to dig deeper into the free page list. */ if (curproc == pageproc && req_class != VM_ALLOC_INTERRUPT) req_class = VM_ALLOC_SYSTEM; /* * Do not allocate reserved pages unless the req has asked for it. */ mtx_lock_flags(&vm_page_queue_free_mtx, MTX_RECURSE); if (vm_cnt.v_free_count + vm_cnt.v_cache_count > vm_cnt.v_free_reserved || (req_class == VM_ALLOC_SYSTEM && vm_cnt.v_free_count + vm_cnt.v_cache_count > vm_cnt.v_interrupt_free_min) || (req_class == VM_ALLOC_INTERRUPT && vm_cnt.v_free_count + vm_cnt.v_cache_count > 0)) m = vm_phys_alloc_freelist_pages(flind, VM_FREEPOOL_DIRECT, 0); else { mtx_unlock(&vm_page_queue_free_mtx); atomic_add_int(&vm_pageout_deficit, max((u_int)req >> VM_ALLOC_COUNT_SHIFT, 1)); pagedaemon_wakeup(); return (NULL); } if (m == NULL) { mtx_unlock(&vm_page_queue_free_mtx); return (NULL); } drop = vm_page_alloc_init(m); mtx_unlock(&vm_page_queue_free_mtx); /* * Initialize the page. Only the PG_ZERO flag is inherited. */ m->aflags = 0; flags = 0; if ((req & VM_ALLOC_ZERO) != 0) flags = PG_ZERO; m->flags &= flags; if ((req & VM_ALLOC_WIRED) != 0) { /* * The page lock is not required for wiring a page that does * not belong to an object. */ atomic_add_int(&vm_cnt.v_wire_count, 1); m->wire_count = 1; } /* Unmanaged pages don't use "act_count". */ m->oflags = VPO_UNMANAGED; if (drop != NULL) vdrop(drop); if (vm_paging_needed()) pagedaemon_wakeup(); return (m); } /* * vm_wait: (also see VM_WAIT macro) * * Sleep until free pages are available for allocation. * - Called in various places before memory allocations. */ void vm_wait(void) { mtx_lock(&vm_page_queue_free_mtx); if (curproc == pageproc) { vm_pageout_pages_needed = 1; msleep(&vm_pageout_pages_needed, &vm_page_queue_free_mtx, PDROP | PSWP, "VMWait", 0); } else { if (!vm_pages_needed) { vm_pages_needed = 1; wakeup(&vm_pages_needed); } msleep(&vm_cnt.v_free_count, &vm_page_queue_free_mtx, PDROP | PVM, "vmwait", 0); } } /* * vm_waitpfault: (also see VM_WAITPFAULT macro) * * Sleep until free pages are available for allocation. * - Called only in vm_fault so that processes page faulting * can be easily tracked. * - Sleeps at a lower priority than vm_wait() so that vm_wait()ing * processes will be able to grab memory first. Do not change * this balance without careful testing first. */ void vm_waitpfault(void) { mtx_lock(&vm_page_queue_free_mtx); if (!vm_pages_needed) { vm_pages_needed = 1; wakeup(&vm_pages_needed); } msleep(&vm_cnt.v_free_count, &vm_page_queue_free_mtx, PDROP | PUSER, "pfault", 0); } struct vm_pagequeue * vm_page_pagequeue(vm_page_t m) { return (&vm_phys_domain(m)->vmd_pagequeues[m->queue]); } /* * vm_page_dequeue: * * Remove the given page from its current page queue. * * The page must be locked. */ void vm_page_dequeue(vm_page_t m) { struct vm_pagequeue *pq; vm_page_assert_locked(m); KASSERT(m->queue < PQ_COUNT, ("vm_page_dequeue: page %p is not queued", m)); pq = vm_page_pagequeue(m); vm_pagequeue_lock(pq); m->queue = PQ_NONE; TAILQ_REMOVE(&pq->pq_pl, m, plinks.q); vm_pagequeue_cnt_dec(pq); vm_pagequeue_unlock(pq); } /* * vm_page_dequeue_locked: * * Remove the given page from its current page queue. * * The page and page queue must be locked. */ void vm_page_dequeue_locked(vm_page_t m) { struct vm_pagequeue *pq; vm_page_lock_assert(m, MA_OWNED); pq = vm_page_pagequeue(m); vm_pagequeue_assert_locked(pq); m->queue = PQ_NONE; TAILQ_REMOVE(&pq->pq_pl, m, plinks.q); vm_pagequeue_cnt_dec(pq); } /* * vm_page_enqueue: * * Add the given page to the specified page queue. * * The page must be locked. */ static void vm_page_enqueue(uint8_t queue, vm_page_t m) { struct vm_pagequeue *pq; vm_page_lock_assert(m, MA_OWNED); KASSERT(queue < PQ_COUNT, ("vm_page_enqueue: invalid queue %u request for page %p", queue, m)); pq = &vm_phys_domain(m)->vmd_pagequeues[queue]; vm_pagequeue_lock(pq); m->queue = queue; TAILQ_INSERT_TAIL(&pq->pq_pl, m, plinks.q); vm_pagequeue_cnt_inc(pq); vm_pagequeue_unlock(pq); } /* * vm_page_requeue: * * Move the given page to the tail of its current page queue. * * The page must be locked. */ void vm_page_requeue(vm_page_t m) { struct vm_pagequeue *pq; vm_page_lock_assert(m, MA_OWNED); KASSERT(m->queue != PQ_NONE, ("vm_page_requeue: page %p is not queued", m)); pq = vm_page_pagequeue(m); vm_pagequeue_lock(pq); TAILQ_REMOVE(&pq->pq_pl, m, plinks.q); TAILQ_INSERT_TAIL(&pq->pq_pl, m, plinks.q); vm_pagequeue_unlock(pq); } /* * vm_page_requeue_locked: * * Move the given page to the tail of its current page queue. * * The page queue must be locked. */ void vm_page_requeue_locked(vm_page_t m) { struct vm_pagequeue *pq; KASSERT(m->queue != PQ_NONE, ("vm_page_requeue_locked: page %p is not queued", m)); pq = vm_page_pagequeue(m); vm_pagequeue_assert_locked(pq); TAILQ_REMOVE(&pq->pq_pl, m, plinks.q); TAILQ_INSERT_TAIL(&pq->pq_pl, m, plinks.q); } /* * vm_page_activate: * * Put the specified page on the active list (if appropriate). * Ensure that act_count is at least ACT_INIT but do not otherwise * mess with it. * * The page must be locked. */ void vm_page_activate(vm_page_t m) { int queue; vm_page_lock_assert(m, MA_OWNED); if ((queue = m->queue) != PQ_ACTIVE) { if (m->wire_count == 0 && (m->oflags & VPO_UNMANAGED) == 0) { if (m->act_count < ACT_INIT) m->act_count = ACT_INIT; if (queue != PQ_NONE) vm_page_dequeue(m); vm_page_enqueue(PQ_ACTIVE, m); } else KASSERT(queue == PQ_NONE, ("vm_page_activate: wired page %p is queued", m)); } else { if (m->act_count < ACT_INIT) m->act_count = ACT_INIT; } } /* * vm_page_free_wakeup: * * Helper routine for vm_page_free_toq() and vm_page_cache(). This * routine is called when a page has been added to the cache or free * queues. * * The page queues must be locked. */ static inline void vm_page_free_wakeup(void) { mtx_assert(&vm_page_queue_free_mtx, MA_OWNED); /* * if pageout daemon needs pages, then tell it that there are * some free. */ if (vm_pageout_pages_needed && vm_cnt.v_cache_count + vm_cnt.v_free_count >= vm_cnt.v_pageout_free_min) { wakeup(&vm_pageout_pages_needed); vm_pageout_pages_needed = 0; } /* * wakeup processes that are waiting on memory if we hit a * high water mark. And wakeup scheduler process if we have * lots of memory. this process will swapin processes. */ if (vm_pages_needed && !vm_page_count_min()) { vm_pages_needed = 0; wakeup(&vm_cnt.v_free_count); } } /* * Turn a cached page into a free page, by changing its attributes. * Keep the statistics up-to-date. * * The free page queue must be locked. */ static void vm_page_cache_turn_free(vm_page_t m) { mtx_assert(&vm_page_queue_free_mtx, MA_OWNED); m->object = NULL; m->valid = 0; KASSERT((m->flags & PG_CACHED) != 0, ("vm_page_cache_turn_free: page %p is not cached", m)); m->flags &= ~PG_CACHED; vm_cnt.v_cache_count--; vm_phys_freecnt_adj(m, 1); } /* * vm_page_free_toq: * * Returns the given page to the free list, * disassociating it with any VM object. * * The object must be locked. The page must be locked if it is managed. */ void vm_page_free_toq(vm_page_t m) { if ((m->oflags & VPO_UNMANAGED) == 0) { vm_page_lock_assert(m, MA_OWNED); KASSERT(!pmap_page_is_mapped(m), ("vm_page_free_toq: freeing mapped page %p", m)); } else KASSERT(m->queue == PQ_NONE, ("vm_page_free_toq: unmanaged page %p is queued", m)); PCPU_INC(cnt.v_tfree); if (vm_page_sbusied(m)) panic("vm_page_free: freeing busy page %p", m); /* * Unqueue, then remove page. Note that we cannot destroy * the page here because we do not want to call the pager's * callback routine until after we've put the page on the * appropriate free queue. */ vm_page_remque(m); vm_page_remove(m); /* * If fictitious remove object association and * return, otherwise delay object association removal. */ if ((m->flags & PG_FICTITIOUS) != 0) { return; } m->valid = 0; vm_page_undirty(m); if (m->wire_count != 0) panic("vm_page_free: freeing wired page %p", m); if (m->hold_count != 0) { m->flags &= ~PG_ZERO; KASSERT((m->flags & PG_UNHOLDFREE) == 0, ("vm_page_free: freeing PG_UNHOLDFREE page %p", m)); m->flags |= PG_UNHOLDFREE; } else { /* * Restore the default memory attribute to the page. */ if (pmap_page_get_memattr(m) != VM_MEMATTR_DEFAULT) pmap_page_set_memattr(m, VM_MEMATTR_DEFAULT); /* * Insert the page into the physical memory allocator's * cache/free page queues. */ mtx_lock(&vm_page_queue_free_mtx); vm_phys_freecnt_adj(m, 1); #if VM_NRESERVLEVEL > 0 if (!vm_reserv_free_page(m)) #else if (TRUE) #endif vm_phys_free_pages(m, 0); if ((m->flags & PG_ZERO) != 0) ++vm_page_zero_count; else vm_page_zero_idle_wakeup(); vm_page_free_wakeup(); mtx_unlock(&vm_page_queue_free_mtx); } } /* * vm_page_wire: * * Mark this page as wired down by yet * another map, removing it from paging queues * as necessary. * * If the page is fictitious, then its wire count must remain one. * * The page must be locked. */ void vm_page_wire(vm_page_t m) { /* * Only bump the wire statistics if the page is not already wired, * and only unqueue the page if it is on some queue (if it is unmanaged * it is already off the queues). */ vm_page_lock_assert(m, MA_OWNED); if ((m->flags & PG_FICTITIOUS) != 0) { KASSERT(m->wire_count == 1, ("vm_page_wire: fictitious page %p's wire count isn't one", m)); return; } if (m->wire_count == 0) { KASSERT((m->oflags & VPO_UNMANAGED) == 0 || m->queue == PQ_NONE, ("vm_page_wire: unmanaged page %p is queued", m)); vm_page_remque(m); atomic_add_int(&vm_cnt.v_wire_count, 1); } m->wire_count++; KASSERT(m->wire_count != 0, ("vm_page_wire: wire_count overflow m=%p", m)); } /* * vm_page_unwire: * * Release one wiring of the specified page, potentially enabling it to be * paged again. If paging is enabled, then the value of the parameter * "queue" determines the queue to which the page is added. * * However, unless the page belongs to an object, it is not enqueued because * it cannot be paged out. * * If a page is fictitious, then its wire count must always be one. * * A managed page must be locked. */ void vm_page_unwire(vm_page_t m, uint8_t queue) { KASSERT(queue < PQ_COUNT, ("vm_page_unwire: invalid queue %u request for page %p", queue, m)); if ((m->oflags & VPO_UNMANAGED) == 0) vm_page_lock_assert(m, MA_OWNED); if ((m->flags & PG_FICTITIOUS) != 0) { KASSERT(m->wire_count == 1, ("vm_page_unwire: fictitious page %p's wire count isn't one", m)); return; } if (m->wire_count > 0) { m->wire_count--; if (m->wire_count == 0) { atomic_subtract_int(&vm_cnt.v_wire_count, 1); if ((m->oflags & VPO_UNMANAGED) != 0 || m->object == NULL) return; if (queue == PQ_INACTIVE) m->flags &= ~PG_WINATCFLS; vm_page_enqueue(queue, m); } } else panic("vm_page_unwire: page %p's wire count is zero", m); } /* * Move the specified page to the inactive queue. * * Many pages placed on the inactive queue should actually go * into the cache, but it is difficult to figure out which. What * we do instead, if the inactive target is well met, is to put * clean pages at the head of the inactive queue instead of the tail. * This will cause them to be moved to the cache more quickly and * if not actively re-referenced, reclaimed more quickly. If we just * stick these pages at the end of the inactive queue, heavy filesystem * meta-data accesses can cause an unnecessary paging load on memory bound * processes. This optimization causes one-time-use metadata to be * reused more quickly. * * Normally athead is 0 resulting in LRU operation. athead is set * to 1 if we want this page to be 'as if it were placed in the cache', * except without unmapping it from the process address space. * * The page must be locked. */ static inline void _vm_page_deactivate(vm_page_t m, int athead) { struct vm_pagequeue *pq; int queue; vm_page_lock_assert(m, MA_OWNED); /* * Ignore if already inactive. */ if ((queue = m->queue) == PQ_INACTIVE) return; if (m->wire_count == 0 && (m->oflags & VPO_UNMANAGED) == 0) { if (queue != PQ_NONE) vm_page_dequeue(m); m->flags &= ~PG_WINATCFLS; pq = &vm_phys_domain(m)->vmd_pagequeues[PQ_INACTIVE]; vm_pagequeue_lock(pq); m->queue = PQ_INACTIVE; if (athead) TAILQ_INSERT_HEAD(&pq->pq_pl, m, plinks.q); else TAILQ_INSERT_TAIL(&pq->pq_pl, m, plinks.q); vm_pagequeue_cnt_inc(pq); vm_pagequeue_unlock(pq); } } /* * Move the specified page to the inactive queue. * * The page must be locked. */ void vm_page_deactivate(vm_page_t m) { _vm_page_deactivate(m, 0); } /* * vm_page_try_to_cache: * * Returns 0 on failure, 1 on success */ int vm_page_try_to_cache(vm_page_t m) { vm_page_lock_assert(m, MA_OWNED); VM_OBJECT_ASSERT_WLOCKED(m->object); if (m->dirty || m->hold_count || m->wire_count || (m->oflags & VPO_UNMANAGED) != 0 || vm_page_busied(m)) return (0); pmap_remove_all(m); if (m->dirty) return (0); vm_page_cache(m); return (1); } /* * vm_page_try_to_free() * * Attempt to free the page. If we cannot free it, we do nothing. * 1 is returned on success, 0 on failure. */ int vm_page_try_to_free(vm_page_t m) { vm_page_lock_assert(m, MA_OWNED); if (m->object != NULL) VM_OBJECT_ASSERT_WLOCKED(m->object); if (m->dirty || m->hold_count || m->wire_count || (m->oflags & VPO_UNMANAGED) != 0 || vm_page_busied(m)) return (0); pmap_remove_all(m); if (m->dirty) return (0); vm_page_free(m); return (1); } /* * vm_page_cache * * Put the specified page onto the page cache queue (if appropriate). * * The object and page must be locked. */ void vm_page_cache(vm_page_t m) { vm_object_t object; boolean_t cache_was_empty; vm_page_lock_assert(m, MA_OWNED); object = m->object; VM_OBJECT_ASSERT_WLOCKED(object); if (vm_page_busied(m) || (m->oflags & VPO_UNMANAGED) || m->hold_count || m->wire_count) panic("vm_page_cache: attempting to cache busy page"); KASSERT(!pmap_page_is_mapped(m), ("vm_page_cache: page %p is mapped", m)); KASSERT(m->dirty == 0, ("vm_page_cache: page %p is dirty", m)); if (m->valid == 0 || object->type == OBJT_DEFAULT || (object->type == OBJT_SWAP && !vm_pager_has_page(object, m->pindex, NULL, NULL))) { /* * Hypothesis: A cache-eligible page belonging to a * default object or swap object but without a backing * store must be zero filled. */ vm_page_free(m); return; } KASSERT((m->flags & PG_CACHED) == 0, ("vm_page_cache: page %p is already cached", m)); /* * Remove the page from the paging queues. */ vm_page_remque(m); /* * Remove the page from the object's collection of resident * pages. */ vm_radix_remove(&object->rtree, m->pindex); TAILQ_REMOVE(&object->memq, m, listq); object->resident_page_count--; /* * Restore the default memory attribute to the page. */ if (pmap_page_get_memattr(m) != VM_MEMATTR_DEFAULT) pmap_page_set_memattr(m, VM_MEMATTR_DEFAULT); /* * Insert the page into the object's collection of cached pages * and the physical memory allocator's cache/free page queues. */ m->flags &= ~PG_ZERO; mtx_lock(&vm_page_queue_free_mtx); cache_was_empty = vm_radix_is_empty(&object->cache); if (vm_radix_insert(&object->cache, m)) { mtx_unlock(&vm_page_queue_free_mtx); if (object->resident_page_count == 0) vdrop(object->handle); m->object = NULL; vm_page_free(m); return; } /* * The above call to vm_radix_insert() could reclaim the one pre- * existing cached page from this object, resulting in a call to * vdrop(). */ if (!cache_was_empty) cache_was_empty = vm_radix_is_singleton(&object->cache); m->flags |= PG_CACHED; vm_cnt.v_cache_count++; PCPU_INC(cnt.v_tcached); #if VM_NRESERVLEVEL > 0 if (!vm_reserv_free_page(m)) { #else if (TRUE) { #endif vm_phys_set_pool(VM_FREEPOOL_CACHE, m, 0); vm_phys_free_pages(m, 0); } vm_page_free_wakeup(); mtx_unlock(&vm_page_queue_free_mtx); /* * Increment the vnode's hold count if this is the object's only * cached page. Decrement the vnode's hold count if this was * the object's only resident page. */ if (object->type == OBJT_VNODE) { if (cache_was_empty && object->resident_page_count != 0) vhold(object->handle); else if (!cache_was_empty && object->resident_page_count == 0) vdrop(object->handle); } } /* * vm_page_advise * * Cache, deactivate, or do nothing as appropriate. This routine * is used by madvise(). * * Generally speaking we want to move the page into the cache so * it gets reused quickly. However, this can result in a silly syndrome * due to the page recycling too quickly. Small objects will not be * fully cached. On the other hand, if we move the page to the inactive * queue we wind up with a problem whereby very large objects * unnecessarily blow away our inactive and cache queues. * * The solution is to move the pages based on a fixed weighting. We * either leave them alone, deactivate them, or move them to the cache, * where moving them to the cache has the highest weighting. * By forcing some pages into other queues we eventually force the * system to balance the queues, potentially recovering other unrelated * space from active. The idea is to not force this to happen too * often. * * The object and page must be locked. */ void vm_page_advise(vm_page_t m, int advice) { int dnw, head; vm_page_assert_locked(m); VM_OBJECT_ASSERT_WLOCKED(m->object); if (advice == MADV_FREE) { /* * Mark the page clean. This will allow the page to be freed * up by the system. However, such pages are often reused * quickly by malloc() so we do not do anything that would * cause a page fault if we can help it. * * Specifically, we do not try to actually free the page now * nor do we try to put it in the cache (which would cause a * page fault on reuse). * * But we do make the page is freeable as we can without * actually taking the step of unmapping it. */ m->dirty = 0; m->act_count = 0; } else if (advice != MADV_DONTNEED) return; dnw = PCPU_GET(dnweight); PCPU_INC(dnweight); /* * Occasionally leave the page alone. */ if ((dnw & 0x01F0) == 0 || m->queue == PQ_INACTIVE) { if (m->act_count >= ACT_INIT) --m->act_count; return; } /* * Clear any references to the page. Otherwise, the page daemon will * immediately reactivate the page. */ vm_page_aflag_clear(m, PGA_REFERENCED); if (advice != MADV_FREE && m->dirty == 0 && pmap_is_modified(m)) vm_page_dirty(m); if (m->dirty || (dnw & 0x0070) == 0) { /* * Deactivate the page 3 times out of 32. */ head = 0; } else { /* * Cache the page 28 times out of every 32. Note that * the page is deactivated instead of cached, but placed * at the head of the queue instead of the tail. */ head = 1; } _vm_page_deactivate(m, head); } /* * Grab a page, waiting until we are waken up due to the page * changing state. We keep on waiting, if the page continues * to be in the object. If the page doesn't exist, first allocate it * and then conditionally zero it. * * This routine may sleep. * * The object must be locked on entry. The lock will, however, be released * and reacquired if the routine sleeps. */ vm_page_t vm_page_grab(vm_object_t object, vm_pindex_t pindex, int allocflags) { vm_page_t m; int sleep; VM_OBJECT_ASSERT_WLOCKED(object); KASSERT((allocflags & VM_ALLOC_SBUSY) == 0 || (allocflags & VM_ALLOC_IGN_SBUSY) != 0, ("vm_page_grab: VM_ALLOC_SBUSY/VM_ALLOC_IGN_SBUSY mismatch")); retrylookup: if ((m = vm_page_lookup(object, pindex)) != NULL) { sleep = (allocflags & VM_ALLOC_IGN_SBUSY) != 0 ? vm_page_xbusied(m) : vm_page_busied(m); if (sleep) { if ((allocflags & VM_ALLOC_NOWAIT) != 0) return (NULL); /* * Reference the page before unlocking and * sleeping so that the page daemon is less * likely to reclaim it. */ vm_page_aflag_set(m, PGA_REFERENCED); vm_page_lock(m); VM_OBJECT_WUNLOCK(object); vm_page_busy_sleep(m, "pgrbwt"); VM_OBJECT_WLOCK(object); goto retrylookup; } else { if ((allocflags & VM_ALLOC_WIRED) != 0) { vm_page_lock(m); vm_page_wire(m); vm_page_unlock(m); } if ((allocflags & (VM_ALLOC_NOBUSY | VM_ALLOC_SBUSY)) == 0) vm_page_xbusy(m); if ((allocflags & VM_ALLOC_SBUSY) != 0) vm_page_sbusy(m); return (m); } } m = vm_page_alloc(object, pindex, allocflags); if (m == NULL) { if ((allocflags & VM_ALLOC_NOWAIT) != 0) return (NULL); VM_OBJECT_WUNLOCK(object); VM_WAIT; VM_OBJECT_WLOCK(object); goto retrylookup; } else if (m->valid != 0) return (m); if (allocflags & VM_ALLOC_ZERO && (m->flags & PG_ZERO) == 0) pmap_zero_page(m); return (m); } /* * Mapping function for valid or dirty bits in a page. * * Inputs are required to range within a page. */ vm_page_bits_t vm_page_bits(int base, int size) { int first_bit; int last_bit; KASSERT( base + size <= PAGE_SIZE, ("vm_page_bits: illegal base/size %d/%d", base, size) ); if (size == 0) /* handle degenerate case */ return (0); first_bit = base >> DEV_BSHIFT; last_bit = (base + size - 1) >> DEV_BSHIFT; return (((vm_page_bits_t)2 << last_bit) - ((vm_page_bits_t)1 << first_bit)); } /* * vm_page_set_valid_range: * * Sets portions of a page valid. The arguments are expected * to be DEV_BSIZE aligned but if they aren't the bitmap is inclusive * of any partial chunks touched by the range. The invalid portion of * such chunks will be zeroed. * * (base + size) must be less then or equal to PAGE_SIZE. */ void vm_page_set_valid_range(vm_page_t m, int base, int size) { int endoff, frag; VM_OBJECT_ASSERT_WLOCKED(m->object); if (size == 0) /* handle degenerate case */ return; /* * If the base is not DEV_BSIZE aligned and the valid * bit is clear, we have to zero out a portion of the * first block. */ if ((frag = base & ~(DEV_BSIZE - 1)) != base && (m->valid & (1 << (base >> DEV_BSHIFT))) == 0) pmap_zero_page_area(m, frag, base - frag); /* * If the ending offset is not DEV_BSIZE aligned and the * valid bit is clear, we have to zero out a portion of * the last block. */ endoff = base + size; if ((frag = endoff & ~(DEV_BSIZE - 1)) != endoff && (m->valid & (1 << (endoff >> DEV_BSHIFT))) == 0) pmap_zero_page_area(m, endoff, DEV_BSIZE - (endoff & (DEV_BSIZE - 1))); /* * Assert that no previously invalid block that is now being validated * is already dirty. */ KASSERT((~m->valid & vm_page_bits(base, size) & m->dirty) == 0, ("vm_page_set_valid_range: page %p is dirty", m)); /* * Set valid bits inclusive of any overlap. */ m->valid |= vm_page_bits(base, size); } /* * Clear the given bits from the specified page's dirty field. */ static __inline void vm_page_clear_dirty_mask(vm_page_t m, vm_page_bits_t pagebits) { uintptr_t addr; #if PAGE_SIZE < 16384 int shift; #endif /* * If the object is locked and the page is neither exclusive busy nor * write mapped, then the page's dirty field cannot possibly be * set by a concurrent pmap operation. */ VM_OBJECT_ASSERT_WLOCKED(m->object); if (!vm_page_xbusied(m) && !pmap_page_is_write_mapped(m)) m->dirty &= ~pagebits; else { /* * The pmap layer can call vm_page_dirty() without * holding a distinguished lock. The combination of * the object's lock and an atomic operation suffice * to guarantee consistency of the page dirty field. * * For PAGE_SIZE == 32768 case, compiler already * properly aligns the dirty field, so no forcible * alignment is needed. Only require existence of * atomic_clear_64 when page size is 32768. */ addr = (uintptr_t)&m->dirty; #if PAGE_SIZE == 32768 atomic_clear_64((uint64_t *)addr, pagebits); #elif PAGE_SIZE == 16384 atomic_clear_32((uint32_t *)addr, pagebits); #else /* PAGE_SIZE <= 8192 */ /* * Use a trick to perform a 32-bit atomic on the * containing aligned word, to not depend on the existence * of atomic_clear_{8, 16}. */ shift = addr & (sizeof(uint32_t) - 1); #if BYTE_ORDER == BIG_ENDIAN shift = (sizeof(uint32_t) - sizeof(m->dirty) - shift) * NBBY; #else shift *= NBBY; #endif addr &= ~(sizeof(uint32_t) - 1); atomic_clear_32((uint32_t *)addr, pagebits << shift); #endif /* PAGE_SIZE */ } } /* * vm_page_set_validclean: * * Sets portions of a page valid and clean. The arguments are expected * to be DEV_BSIZE aligned but if they aren't the bitmap is inclusive * of any partial chunks touched by the range. The invalid portion of * such chunks will be zero'd. * * (base + size) must be less then or equal to PAGE_SIZE. */ void vm_page_set_validclean(vm_page_t m, int base, int size) { vm_page_bits_t oldvalid, pagebits; int endoff, frag; VM_OBJECT_ASSERT_WLOCKED(m->object); if (size == 0) /* handle degenerate case */ return; /* * If the base is not DEV_BSIZE aligned and the valid * bit is clear, we have to zero out a portion of the * first block. */ if ((frag = base & ~(DEV_BSIZE - 1)) != base && (m->valid & ((vm_page_bits_t)1 << (base >> DEV_BSHIFT))) == 0) pmap_zero_page_area(m, frag, base - frag); /* * If the ending offset is not DEV_BSIZE aligned and the * valid bit is clear, we have to zero out a portion of * the last block. */ endoff = base + size; if ((frag = endoff & ~(DEV_BSIZE - 1)) != endoff && (m->valid & ((vm_page_bits_t)1 << (endoff >> DEV_BSHIFT))) == 0) pmap_zero_page_area(m, endoff, DEV_BSIZE - (endoff & (DEV_BSIZE - 1))); /* * Set valid, clear dirty bits. If validating the entire * page we can safely clear the pmap modify bit. We also * use this opportunity to clear the VPO_NOSYNC flag. If a process * takes a write fault on a MAP_NOSYNC memory area the flag will * be set again. * * We set valid bits inclusive of any overlap, but we can only * clear dirty bits for DEV_BSIZE chunks that are fully within * the range. */ oldvalid = m->valid; pagebits = vm_page_bits(base, size); m->valid |= pagebits; #if 0 /* NOT YET */ if ((frag = base & (DEV_BSIZE - 1)) != 0) { frag = DEV_BSIZE - frag; base += frag; size -= frag; if (size < 0) size = 0; } pagebits = vm_page_bits(base, size & (DEV_BSIZE - 1)); #endif if (base == 0 && size == PAGE_SIZE) { /* * The page can only be modified within the pmap if it is * mapped, and it can only be mapped if it was previously * fully valid. */ if (oldvalid == VM_PAGE_BITS_ALL) /* * Perform the pmap_clear_modify() first. Otherwise, * a concurrent pmap operation, such as * pmap_protect(), could clear a modification in the * pmap and set the dirty field on the page before * pmap_clear_modify() had begun and after the dirty * field was cleared here. */ pmap_clear_modify(m); m->dirty = 0; m->oflags &= ~VPO_NOSYNC; } else if (oldvalid != VM_PAGE_BITS_ALL) m->dirty &= ~pagebits; else vm_page_clear_dirty_mask(m, pagebits); } void vm_page_clear_dirty(vm_page_t m, int base, int size) { vm_page_clear_dirty_mask(m, vm_page_bits(base, size)); } /* * vm_page_set_invalid: * * Invalidates DEV_BSIZE'd chunks within a page. Both the * valid and dirty bits for the effected areas are cleared. */ void vm_page_set_invalid(vm_page_t m, int base, int size) { vm_page_bits_t bits; vm_object_t object; object = m->object; VM_OBJECT_ASSERT_WLOCKED(object); if (object->type == OBJT_VNODE && base == 0 && IDX_TO_OFF(m->pindex) + size >= object->un_pager.vnp.vnp_size) bits = VM_PAGE_BITS_ALL; else bits = vm_page_bits(base, size); if (m->valid == VM_PAGE_BITS_ALL && bits != 0) pmap_remove_all(m); KASSERT((bits == 0 && m->valid == VM_PAGE_BITS_ALL) || !pmap_page_is_mapped(m), ("vm_page_set_invalid: page %p is mapped", m)); m->valid &= ~bits; m->dirty &= ~bits; } /* * vm_page_zero_invalid() * * The kernel assumes that the invalid portions of a page contain * garbage, but such pages can be mapped into memory by user code. * When this occurs, we must zero out the non-valid portions of the * page so user code sees what it expects. * * Pages are most often semi-valid when the end of a file is mapped * into memory and the file's size is not page aligned. */ void vm_page_zero_invalid(vm_page_t m, boolean_t setvalid) { int b; int i; VM_OBJECT_ASSERT_WLOCKED(m->object); /* * Scan the valid bits looking for invalid sections that * must be zeroed. Invalid sub-DEV_BSIZE'd areas ( where the * valid bit may be set ) have already been zeroed by * vm_page_set_validclean(). */ for (b = i = 0; i <= PAGE_SIZE / DEV_BSIZE; ++i) { if (i == (PAGE_SIZE / DEV_BSIZE) || (m->valid & ((vm_page_bits_t)1 << i))) { if (i > b) { pmap_zero_page_area(m, b << DEV_BSHIFT, (i - b) << DEV_BSHIFT); } b = i + 1; } } /* * setvalid is TRUE when we can safely set the zero'd areas * as being valid. We can do this if there are no cache consistancy * issues. e.g. it is ok to do with UFS, but not ok to do with NFS. */ if (setvalid) m->valid = VM_PAGE_BITS_ALL; } /* * vm_page_is_valid: * * Is (partial) page valid? Note that the case where size == 0 * will return FALSE in the degenerate case where the page is * entirely invalid, and TRUE otherwise. */ int vm_page_is_valid(vm_page_t m, int base, int size) { vm_page_bits_t bits; VM_OBJECT_ASSERT_LOCKED(m->object); bits = vm_page_bits(base, size); return (m->valid != 0 && (m->valid & bits) == bits); } /* * vm_page_ps_is_valid: * * Returns TRUE if the entire (super)page is valid and FALSE otherwise. */ boolean_t vm_page_ps_is_valid(vm_page_t m) { int i, npages; VM_OBJECT_ASSERT_LOCKED(m->object); npages = atop(pagesizes[m->psind]); /* * The physically contiguous pages that make up a superpage, i.e., a * page with a page size index ("psind") greater than zero, will * occupy adjacent entries in vm_page_array[]. */ for (i = 0; i < npages; i++) { if (m[i].valid != VM_PAGE_BITS_ALL) return (FALSE); } return (TRUE); } /* * Set the page's dirty bits if the page is modified. */ void vm_page_test_dirty(vm_page_t m) { VM_OBJECT_ASSERT_WLOCKED(m->object); if (m->dirty != VM_PAGE_BITS_ALL && pmap_is_modified(m)) vm_page_dirty(m); } void vm_page_lock_KBI(vm_page_t m, const char *file, int line) { mtx_lock_flags_(vm_page_lockptr(m), 0, file, line); } void vm_page_unlock_KBI(vm_page_t m, const char *file, int line) { mtx_unlock_flags_(vm_page_lockptr(m), 0, file, line); } int vm_page_trylock_KBI(vm_page_t m, const char *file, int line) { return (mtx_trylock_flags_(vm_page_lockptr(m), 0, file, line)); } #if defined(INVARIANTS) || defined(INVARIANT_SUPPORT) void vm_page_assert_locked_KBI(vm_page_t m, const char *file, int line) { vm_page_lock_assert_KBI(m, MA_OWNED, file, line); } void vm_page_lock_assert_KBI(vm_page_t m, int a, const char *file, int line) { mtx_assert_(vm_page_lockptr(m), a, file, line); } #endif #ifdef INVARIANTS void vm_page_object_lock_assert(vm_page_t m) { /* * Certain of the page's fields may only be modified by the * holder of the containing object's lock or the exclusive busy. * holder. Unfortunately, the holder of the write busy is * not recorded, and thus cannot be checked here. */ if (m->object != NULL && !vm_page_xbusied(m)) VM_OBJECT_ASSERT_WLOCKED(m->object); } void vm_page_assert_pga_writeable(vm_page_t m, uint8_t bits) { if ((bits & PGA_WRITEABLE) == 0) return; /* * The PGA_WRITEABLE flag can only be set if the page is * managed, is exclusively busied or the object is locked. * Currently, this flag is only set by pmap_enter(). */ KASSERT((m->oflags & VPO_UNMANAGED) == 0, ("PGA_WRITEABLE on unmanaged page")); if (!vm_page_xbusied(m)) VM_OBJECT_ASSERT_LOCKED(m->object); } #endif #include "opt_ddb.h" #ifdef DDB #include #include DB_SHOW_COMMAND(page, vm_page_print_page_info) { db_printf("vm_cnt.v_free_count: %d\n", vm_cnt.v_free_count); db_printf("vm_cnt.v_cache_count: %d\n", vm_cnt.v_cache_count); db_printf("vm_cnt.v_inactive_count: %d\n", vm_cnt.v_inactive_count); db_printf("vm_cnt.v_active_count: %d\n", vm_cnt.v_active_count); db_printf("vm_cnt.v_wire_count: %d\n", vm_cnt.v_wire_count); db_printf("vm_cnt.v_free_reserved: %d\n", vm_cnt.v_free_reserved); db_printf("vm_cnt.v_free_min: %d\n", vm_cnt.v_free_min); db_printf("vm_cnt.v_free_target: %d\n", vm_cnt.v_free_target); db_printf("vm_cnt.v_cache_min: %d\n", vm_cnt.v_cache_min); db_printf("vm_cnt.v_inactive_target: %d\n", vm_cnt.v_inactive_target); } DB_SHOW_COMMAND(pageq, vm_page_print_pageq_info) { int dom; db_printf("pq_free %d pq_cache %d\n", vm_cnt.v_free_count, vm_cnt.v_cache_count); for (dom = 0; dom < vm_ndomains; dom++) { db_printf( "dom %d page_cnt %d free %d pq_act %d pq_inact %d pass %d\n", dom, vm_dom[dom].vmd_page_count, vm_dom[dom].vmd_free_count, vm_dom[dom].vmd_pagequeues[PQ_ACTIVE].pq_cnt, vm_dom[dom].vmd_pagequeues[PQ_INACTIVE].pq_cnt, vm_dom[dom].vmd_pass); } } DB_SHOW_COMMAND(pginfo, vm_page_print_pginfo) { vm_page_t m; boolean_t phys; if (!have_addr) { db_printf("show pginfo addr\n"); return; } phys = strchr(modif, 'p') != NULL; if (phys) m = PHYS_TO_VM_PAGE(addr); else m = (vm_page_t)addr; db_printf( "page %p obj %p pidx 0x%jx phys 0x%jx q %d hold %d wire %d\n" " af 0x%x of 0x%x f 0x%x act %d busy %x valid 0x%x dirty 0x%x\n", m, m->object, (uintmax_t)m->pindex, (uintmax_t)m->phys_addr, m->queue, m->hold_count, m->wire_count, m->aflags, m->oflags, m->flags, m->act_count, m->busy_lock, m->valid, m->dirty); } #endif /* DDB */ Index: projects/clang-trunk/sys =================================================================== --- projects/clang-trunk/sys (revision 283870) +++ projects/clang-trunk/sys (revision 283871) Property changes on: projects/clang-trunk/sys ___________________________________________________________________ Modified: svn:mergeinfo ## -0,0 +0,1 ## Merged /head/sys:r283771-283870 Index: projects/clang-trunk/tests/Makefile =================================================================== --- projects/clang-trunk/tests/Makefile (revision 283870) +++ projects/clang-trunk/tests/Makefile (revision 283871) @@ -1,15 +1,15 @@ # $FreeBSD$ .include -SUBDIR= etc -SUBDIR= sys +SUBDIR+= etc +SUBDIR+= sys TESTSDIR= ${TESTSBASE} KYUAFILE= yes afterinstall: install-tests-local install-tests-local: .PHONY ${INSTALL_SYMLINK} ../local/tests ${DESTDIR}${TESTSDIR}/local .include Index: projects/clang-trunk/tests/sys/kern/ptrace_test.c =================================================================== --- projects/clang-trunk/tests/sys/kern/ptrace_test.c (revision 283870) +++ projects/clang-trunk/tests/sys/kern/ptrace_test.c (revision 283871) @@ -1,389 +1,391 @@ /*- * Copyright (c) 2015 John Baldwin * All rights reserved. * * Redistribution and use in source and binary forms, with or without * modification, are permitted provided that the following conditions * are met: * 1. Redistributions of source code must retain the above copyright * notice, this list of conditions and the following disclaimer. * 2. Redistributions in binary form must reproduce the above copyright * notice, this list of conditions and the following disclaimer in the * documentation and/or other materials provided with the distribution. * * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE * ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF * SUCH DAMAGE. */ #include __FBSDID("$FreeBSD$"); #include #include #include #include #include #include #include #include #include #include /* * Verify that a parent debugger process "sees" the exit of a debugged * process exactly once when attached via PT_TRACE_ME. */ ATF_TC_WITHOUT_HEAD(ptrace__parent_wait_after_trace_me); ATF_TC_BODY(ptrace__parent_wait_after_trace_me, tc) { pid_t child, wpid; int status; ATF_REQUIRE((child = fork()) != -1); if (child == 0) { /* Child process. */ ATF_REQUIRE(ptrace(PT_TRACE_ME, 0, NULL, 0) != -1); /* Trigger a stop. */ raise(SIGSTOP); exit(1); } /* Parent process. */ /* The first wait() should report the stop from SIGSTOP. */ wpid = waitpid(child, &status, 0); ATF_REQUIRE(wpid == child); ATF_REQUIRE(WIFSTOPPED(status)); ATF_REQUIRE(WSTOPSIG(status) == SIGSTOP); /* Continue the child ignoring the SIGSTOP. */ ATF_REQUIRE(ptrace(PT_CONTINUE, child, (caddr_t)1, 0) != -1); /* The second wait() should report the exit status. */ wpid = waitpid(child, &status, 0); ATF_REQUIRE(wpid == child); ATF_REQUIRE(WIFEXITED(status)); ATF_REQUIRE(WEXITSTATUS(status) == 1); /* The child should no longer exist. */ wpid = waitpid(child, &status, 0); ATF_REQUIRE(wpid == -1); ATF_REQUIRE(errno == ECHILD); } /* * Verify that a parent debugger process "sees" the exit of a debugged * process exactly once when attached via PT_ATTACH. */ ATF_TC_WITHOUT_HEAD(ptrace__parent_wait_after_attach); ATF_TC_BODY(ptrace__parent_wait_after_attach, tc) { pid_t child, wpid; int cpipe[2], status; char c; ATF_REQUIRE(pipe(cpipe) == 0); ATF_REQUIRE((child = fork()) != -1); if (child == 0) { /* Child process. */ close(cpipe[0]); /* Wait for the parent to attach. */ ATF_REQUIRE(read(cpipe[1], &c, sizeof(c)) == 0); exit(1); } close(cpipe[1]); /* Parent process. */ /* Attach to the child process. */ ATF_REQUIRE(ptrace(PT_ATTACH, child, NULL, 0) == 0); /* The first wait() should report the SIGSTOP from PT_ATTACH. */ wpid = waitpid(child, &status, 0); ATF_REQUIRE(wpid == child); ATF_REQUIRE(WIFSTOPPED(status)); ATF_REQUIRE(WSTOPSIG(status) == SIGSTOP); /* Continue the child ignoring the SIGSTOP. */ ATF_REQUIRE(ptrace(PT_CONTINUE, child, (caddr_t)1, 0) != -1); /* Signal the child to exit. */ close(cpipe[0]); /* The second wait() should report the exit status. */ wpid = waitpid(child, &status, 0); ATF_REQUIRE(wpid == child); ATF_REQUIRE(WIFEXITED(status)); ATF_REQUIRE(WEXITSTATUS(status) == 1); /* The child should no longer exist. */ wpid = waitpid(child, &status, 0); ATF_REQUIRE(wpid == -1); ATF_REQUIRE(errno == ECHILD); } /* * Verify that a parent process "sees" the exit of a debugged process only * after the debugger has seen it. */ ATF_TC_WITHOUT_HEAD(ptrace__parent_sees_exit_after_child_debugger); ATF_TC_BODY(ptrace__parent_sees_exit_after_child_debugger, tc) { pid_t child, debugger, wpid; int cpipe[2], dpipe[2], status; char c; ATF_REQUIRE(pipe(cpipe) == 0); ATF_REQUIRE((child = fork()) != -1); if (child == 0) { /* Child process. */ close(cpipe[0]); /* Wait for parent to be ready. */ ATF_REQUIRE(read(cpipe[1], &c, sizeof(c)) == sizeof(c)); exit(1); } close(cpipe[1]); ATF_REQUIRE(pipe(dpipe) == 0); ATF_REQUIRE((debugger = fork()) != -1); if (debugger == 0) { /* Debugger process. */ close(dpipe[0]); ATF_REQUIRE(ptrace(PT_ATTACH, child, NULL, 0) != -1); wpid = waitpid(child, &status, 0); ATF_REQUIRE(wpid == child); ATF_REQUIRE(WIFSTOPPED(status)); ATF_REQUIRE(WSTOPSIG(status) == SIGSTOP); ATF_REQUIRE(ptrace(PT_CONTINUE, child, (caddr_t)1, 0) != -1); /* Signal parent that debugger is attached. */ ATF_REQUIRE(write(dpipe[1], &c, sizeof(c)) == sizeof(c)); /* Wait for parent's failed wait. */ ATF_REQUIRE(read(dpipe[1], &c, sizeof(c)) == 0); wpid = waitpid(child, &status, 0); ATF_REQUIRE(wpid == child); ATF_REQUIRE(WIFEXITED(status)); ATF_REQUIRE(WEXITSTATUS(status) == 1); exit(0); } close(dpipe[1]); /* Parent process. */ /* Wait for the debugger to attach to the child. */ ATF_REQUIRE(read(dpipe[0], &c, sizeof(c)) == sizeof(c)); /* Release the child. */ ATF_REQUIRE(write(cpipe[0], &c, sizeof(c)) == sizeof(c)); ATF_REQUIRE(read(cpipe[0], &c, sizeof(c)) == 0); close(cpipe[0]); /* * Wait for the child to exit. This is kind of gross, but * there is not a better way. */ for (;;) { struct kinfo_proc kp; size_t len; int mib[4]; mib[0] = CTL_KERN; mib[1] = KERN_PROC; mib[2] = KERN_PROC_PID; mib[3] = child; len = sizeof(kp); if (sysctl(mib, nitems(mib), &kp, &len, NULL, 0) == -1) { /* The KERN_PROC_PID sysctl fails for zombies. */ ATF_REQUIRE(errno == ESRCH); break; } usleep(5000); } /* * This wait should return a pid of 0 to indicate no status to * report. The parent should see the child as non-exited * until the debugger sees the exit. */ wpid = waitpid(child, &status, WNOHANG); ATF_REQUIRE(wpid == 0); /* Signal the debugger to wait for the child. */ close(dpipe[0]); /* Wait for the debugger. */ wpid = waitpid(debugger, &status, 0); ATF_REQUIRE(wpid == debugger); ATF_REQUIRE(WIFEXITED(status)); ATF_REQUIRE(WEXITSTATUS(status) == 0); /* The child process should now be ready. */ wpid = waitpid(child, &status, WNOHANG); ATF_REQUIRE(wpid == child); ATF_REQUIRE(WIFEXITED(status)); ATF_REQUIRE(WEXITSTATUS(status) == 1); } /* * Verify that a parent process "sees" the exit of a debugged process * only after a non-direct-child debugger has seen it. In particular, * various wait() calls in the parent must avoid failing with ESRCH by * checking the parent's orphan list for the debugee. */ ATF_TC_WITHOUT_HEAD(ptrace__parent_sees_exit_after_unrelated_debugger); ATF_TC_BODY(ptrace__parent_sees_exit_after_unrelated_debugger, tc) { pid_t child, debugger, fpid, wpid; int cpipe[2], dpipe[2], status; char c; ATF_REQUIRE(pipe(cpipe) == 0); ATF_REQUIRE((child = fork()) != -1); if (child == 0) { /* Child process. */ close(cpipe[0]); /* Wait for parent to be ready. */ ATF_REQUIRE(read(cpipe[1], &c, sizeof(c)) == sizeof(c)); exit(1); } close(cpipe[1]); ATF_REQUIRE(pipe(dpipe) == 0); ATF_REQUIRE((debugger = fork()) != -1); if (debugger == 0) { /* Debugger parent. */ /* * Fork again and drop the debugger parent so that the * debugger is not a child of the main parent. */ ATF_REQUIRE((fpid = fork()) != -1); if (fpid != 0) exit(2); /* Debugger process. */ close(dpipe[0]); ATF_REQUIRE(ptrace(PT_ATTACH, child, NULL, 0) != -1); wpid = waitpid(child, &status, 0); ATF_REQUIRE(wpid == child); ATF_REQUIRE(WIFSTOPPED(status)); ATF_REQUIRE(WSTOPSIG(status) == SIGSTOP); ATF_REQUIRE(ptrace(PT_CONTINUE, child, (caddr_t)1, 0) != -1); /* Signal parent that debugger is attached. */ ATF_REQUIRE(write(dpipe[1], &c, sizeof(c)) == sizeof(c)); /* Wait for parent's failed wait. */ - ATF_REQUIRE(read(dpipe[1], &c, sizeof(c)) == 0); + ATF_REQUIRE(read(dpipe[1], &c, sizeof(c)) == sizeof(c)); wpid = waitpid(child, &status, 0); ATF_REQUIRE(wpid == child); ATF_REQUIRE(WIFEXITED(status)); ATF_REQUIRE(WEXITSTATUS(status) == 1); exit(0); } + close(dpipe[1]); /* Parent process. */ /* Wait for the debugger parent process to exit. */ wpid = waitpid(debugger, &status, 0); ATF_REQUIRE(wpid == debugger); ATF_REQUIRE(WIFEXITED(status)); ATF_REQUIRE(WEXITSTATUS(status) == 2); /* A WNOHANG wait here should see the non-exited child. */ wpid = waitpid(child, &status, WNOHANG); ATF_REQUIRE(wpid == 0); /* Wait for the debugger to attach to the child. */ ATF_REQUIRE(read(dpipe[0], &c, sizeof(c)) == sizeof(c)); /* Release the child. */ ATF_REQUIRE(write(cpipe[0], &c, sizeof(c)) == sizeof(c)); ATF_REQUIRE(read(cpipe[0], &c, sizeof(c)) == 0); close(cpipe[0]); /* * Wait for the child to exit. This is kind of gross, but * there is not a better way. */ for (;;) { struct kinfo_proc kp; size_t len; int mib[4]; mib[0] = CTL_KERN; mib[1] = KERN_PROC; mib[2] = KERN_PROC_PID; mib[3] = child; len = sizeof(kp); if (sysctl(mib, nitems(mib), &kp, &len, NULL, 0) == -1) { /* The KERN_PROC_PID sysctl fails for zombies. */ ATF_REQUIRE(errno == ESRCH); break; } usleep(5000); } /* * This wait should return a pid of 0 to indicate no status to * report. The parent should see the child as non-exited * until the debugger sees the exit. */ wpid = waitpid(child, &status, WNOHANG); ATF_REQUIRE(wpid == 0); /* Signal the debugger to wait for the child. */ - close(dpipe[0]); + ATF_REQUIRE(write(dpipe[0], &c, sizeof(c)) == sizeof(c)); /* Wait for the debugger. */ - ATF_REQUIRE(read(dpipe[1], &c, sizeof(c)) == 0); + ATF_REQUIRE(read(dpipe[0], &c, sizeof(c)) == 0); + close(dpipe[0]); /* The child process should now be ready. */ wpid = waitpid(child, &status, WNOHANG); ATF_REQUIRE(wpid == child); ATF_REQUIRE(WIFEXITED(status)); ATF_REQUIRE(WEXITSTATUS(status) == 1); } ATF_TP_ADD_TCS(tp) { ATF_TP_ADD_TC(tp, ptrace__parent_wait_after_trace_me); ATF_TP_ADD_TC(tp, ptrace__parent_wait_after_attach); ATF_TP_ADD_TC(tp, ptrace__parent_sees_exit_after_child_debugger); ATF_TP_ADD_TC(tp, ptrace__parent_sees_exit_after_unrelated_debugger); return (atf_no_error()); } Index: projects/clang-trunk/tools/build/options/WITHOUT_MANDOCDB =================================================================== --- projects/clang-trunk/tools/build/options/WITHOUT_MANDOCDB (nonexistent) +++ projects/clang-trunk/tools/build/options/WITHOUT_MANDOCDB (revision 283871) @@ -0,0 +1,6 @@ +.\" $FreeBSD$ +Use the +.Xr mandoc 1 +version of +.Xr makewhatis 8 +database and utilities. Property changes on: projects/clang-trunk/tools/build/options/WITHOUT_MANDOCDB ___________________________________________________________________ Added: svn:eol-style ## -0,0 +1 ## +native \ No newline at end of property Added: svn:keywords ## -0,0 +1 ## +FreeBSD=%H \ No newline at end of property Added: svn:mime-type ## -0,0 +1 ## +text/plain \ No newline at end of property Index: projects/clang-trunk/usr.bin/Makefile =================================================================== --- projects/clang-trunk/usr.bin/Makefile (revision 283870) +++ projects/clang-trunk/usr.bin/Makefile (revision 283871) @@ -1,432 +1,434 @@ # From: @(#)Makefile 8.3 (Berkeley) 1/7/94 # $FreeBSD$ .include # XXX MISSING: deroff diction graph learn plot # spell spline struct xsend # XXX Use GNU versions: diff ld patch # Moved to secure: bdes # SUBDIR= ${_addr2line} \ alias \ apply \ asa \ awk \ banner \ basename \ brandelf \ bsdiff \ bzip2 \ bzip2recover \ cap_mkdb \ chat \ chpass \ cksum \ ${_clang} \ cmp \ col \ colldef \ colrm \ column \ comm \ compress \ cpuset \ csplit \ ctlstat \ cut \ ${_cxxfilt} \ dirname \ dpv \ du \ elf2aout \ ${_elfcopy} \ elfdump \ enigma \ env \ expand \ false \ fetch \ find \ fmt \ fold \ fstat \ fsync \ gcore \ gencat \ getconf \ getent \ getopt \ grep \ gzip \ head \ hexdump \ ${_iconv} \ id \ ipcrm \ ipcs \ join \ jot \ ${_kdump} \ keylogin \ keylogout \ killall \ ktrace \ ktrdump \ lam \ lastcomm \ ldd \ leave \ less \ lessecho \ lesskey \ limits \ locale \ lock \ lockf \ logger \ login \ logins \ logname \ look \ lorder \ lsvfs \ lzmainfo \ m4 \ ${_makewhatis} \ ${_man} \ mandoc \ mesg \ minigzip \ ministat \ ${_mkcsmapper} \ mkdep \ ${_mkesdb} \ mkfifo \ mkimg \ mklocale \ mktemp \ mkulzma \ mkuzip \ mt \ ncal \ netstat \ newgrp \ nfsstat \ nice \ nl \ ${_nm} \ nohup \ opieinfo \ opiekey \ opiepasswd \ pagesize \ passwd \ paste \ patch \ pathchk \ perror \ pr \ printenv \ printf \ procstat \ protect \ rctl \ ${_readelf} \ renice \ rev \ revoke \ rpcinfo \ rs \ rup \ rusers \ rwall \ script \ sed \ send-pr \ seq \ shar \ showmount \ ${_size} \ sockstat \ soelim \ sort \ split \ stat \ stdbuf \ ${_strings} \ su \ systat \ tabs \ tail \ tar \ tcopy \ tee \ ${_tests} \ time \ timeout \ tip \ top \ touch \ tput \ tr \ true \ truncate \ ${_truss} \ tset \ tsort \ tty \ uname \ unexpand \ uniq \ unzip \ units \ unvis \ uudecode \ uuencode \ vis \ vmstat \ w \ wall \ wc \ what \ whereis \ which \ whois \ write \ xargs \ xinstall \ xo \ xz \ xzdec \ yes \ ${_ypcat} \ ${_ypmatch} \ ${_ypwhich} # NB: keep these sorted by MK_* knobs .if ${MK_AT} != "no" SUBDIR+= at .endif .if ${MK_ATM} != "no" SUBDIR+= atm .endif .if ${MK_BLUETOOTH} != "no" SUBDIR+= bluetooth .endif .if ${MK_BSD_CPIO} != "no" SUBDIR+= cpio .endif .if ${MK_CALENDAR} != "no" SUBDIR+= calendar .endif .if ${MK_CLANG} != "no" _clang= clang .endif .if ${MK_EE} != "no" SUBDIR+= ee .endif .if ${MK_ELFTOOLCHAIN_TOOLS} != "no" _addr2line= addr2line _cxxfilt= cxxfilt _elfcopy= elfcopy _nm= nm _readelf= readelf _size= size _strings= strings .endif .if ${MK_FILE} != "no" SUBDIR+= file .endif .if ${MK_FINGER} != "no" SUBDIR+= finger .endif .if ${MK_FMAKE} != "no" SUBDIR+= make .endif .if ${MK_FTP} != "no" SUBDIR+= ftp .endif .if ${MK_GPL_DTC} != "yes" SUBDIR+= dtc .endif .if ${MK_GROFF} != "no" SUBDIR+= vgrind .endif .if ${MK_HESIOD} != "no" SUBDIR+= hesinfo .endif .if ${MK_ICONV} != "no" _iconv= iconv _mkcsmapper= mkcsmapper _mkesdb= mkesdb .endif .if ${MK_ISCSI} != "no" SUBDIR+= iscsictl .endif .if ${MK_KDUMP} != "no" SUBDIR+= kdump .if ${MACHINE_ARCH} != "aarch64" # ARM64TODO truss does not build SUBDIR+= truss .endif .endif .if ${MK_KERBEROS_SUPPORT} != "no" SUBDIR+= compile_et .endif .if ${MK_LDNS_UTILS} != "no" SUBDIR+= drill SUBDIR+= host .endif .if ${MK_LOCATE} != "no" SUBDIR+= locate .endif # XXX msgs? .if ${MK_MAIL} != "no" SUBDIR+= biff SUBDIR+= from SUBDIR+= mail SUBDIR+= msgs .endif .if ${MK_MAKE} != "no" SUBDIR+= bmake .endif .if ${MK_MAN_UTILS} != "no" SUBDIR+= catman +.if ${MK_MANDOCDB} == "no" _makewhatis= makewhatis +.endif _man= man .endif .if ${MK_NETCAT} != "no" SUBDIR+= nc .endif .if ${MK_NIS} != "no" SUBDIR+= ypcat SUBDIR+= ypmatch SUBDIR+= ypwhich .endif .if ${MK_OPENSSH} != "no" SUBDIR+= ssh-copy-id .endif .if ${MK_OPENSSL} != "no" SUBDIR+= bc SUBDIR+= chkey SUBDIR+= dc SUBDIR+= newkey .endif .if ${MK_QUOTAS} != "no" SUBDIR+= quota .endif .if ${MK_RCMDS} != "no" SUBDIR+= rlogin SUBDIR+= rsh SUBDIR+= ruptime SUBDIR+= rwho .endif .if ${MK_SENDMAIL} != "no" SUBDIR+= vacation .endif .if ${MK_TALK} != "no" SUBDIR+= talk .endif .if ${MK_TELNET} != "no" SUBDIR+= telnet .endif .if ${MK_TESTS} != "no" _tests= tests .endif .if ${MK_TEXTPROC} != "no" SUBDIR+= checknr SUBDIR+= colcrt SUBDIR+= ul .endif .if ${MK_TFTP} != "no" SUBDIR+= tftp .endif .if ${MK_TOOLCHAIN} != "no" SUBDIR+= ar SUBDIR+= c89 SUBDIR+= c99 SUBDIR+= ctags SUBDIR+= file2c .if ${MACHINE_ARCH} != "aarch64" # ARM64TODO gprof does not build SUBDIR+= gprof .endif SUBDIR+= indent SUBDIR+= lex SUBDIR+= mkstr SUBDIR+= rpcgen SUBDIR+= unifdef .if ${MACHINE_ARCH} != "aarch64" # ARM64TODO xlint does not build SUBDIR+= xlint .endif SUBDIR+= xstr SUBDIR+= yacc .endif .if ${MK_VI} != "no" SUBDIR+= vi .endif .if ${MK_VT} != "no" SUBDIR+= vtfontcvt .endif .if ${MK_USB} != "no" SUBDIR+= usbhidaction SUBDIR+= usbhidctl .endif .if ${MK_UTMPX} != "no" SUBDIR+= last SUBDIR+= users SUBDIR+= who .endif .if ${MK_SVN} == "yes" || ${MK_SVNLITE} == "yes" SUBDIR+= svn .endif .include SUBDIR:= ${SUBDIR:O} SUBDIR_PARALLEL= .include Index: projects/clang-trunk/usr.bin/man/Makefile =================================================================== --- projects/clang-trunk/usr.bin/man/Makefile (revision 283870) +++ projects/clang-trunk/usr.bin/man/Makefile (revision 283871) @@ -1,11 +1,17 @@ # $FreeBSD$ +.include + SCRIPTS= man.sh -LINKS= ${BINDIR}/man ${BINDIR}/apropos \ - ${BINDIR}/man ${BINDIR}/manpath \ - ${BINDIR}/man ${BINDIR}/whatis +LINKS= ${BINDIR}/man ${BINDIR}/manpath -MAN= man.1 manpath.1 apropos.1 man.conf.5 +MAN= man.1 manpath.1 man.conf.5 + +.if ${MK_MANDOCDB} == no +LINKS+= ${BINDIR}/man ${BINDIR}/apropos \ + ${BINDIR}/man ${BINDIR}/whatis +MAN+= apropos.1 MLINKS= apropos.1 whatis.1 +.endif .include Index: projects/clang-trunk/usr.bin/mandoc/Makefile =================================================================== --- projects/clang-trunk/usr.bin/mandoc/Makefile (revision 283870) +++ projects/clang-trunk/usr.bin/mandoc/Makefile (revision 283871) @@ -1,80 +1,89 @@ # $FreeBSD$ +.include + MDOCMLDIR= ${.CURDIR}/../../contrib/mdocml .PATH: ${MDOCMLDIR} PROG= mandoc FILES= example.style.css style.css FILESDIR= ${SHAREDIR}/mdocml MAN= mandoc.1 eqn.7 mandoc_char.7 tbl.7 man.7 mdoc.7 # roff.7 -MLINKS= mandoc.1 mdocml.1 +MLINKS= mandoc.1 mdocml.1 +.if ${MK_MANDOCDB} != no +MAN+= apropos.1 makewhatis.8 +MLINKS+= apropos.1 whatis.1 +LINKS= ${BINDIR}/mandoc ${BINDIR}/whatis \ + ${BINDIR}/mandoc ${BINDIR}/makewhatis \ + ${BINDIR}/mandoc ${BINDIR}/apropos +.endif LIBMAN_SRCS= man.c \ man_hash.c \ man_macro.c \ man_validate.c LIBMDOC_SRCS= att.c \ lib.c \ mdoc.c \ mdoc_argv.c \ mdoc_hash.c \ mdoc_macro.c \ mdoc_validate.c \ st.c \ LIBROFF_SRCS= eqn.c \ roff.c \ tbl.c \ tbl_data.c \ tbl_layout.c \ tbl_opts.c \ LIB_SRCS= ${LIBMAN_SRCS} \ ${LIBMDOC_SRCS} \ ${LIBROFF_SRCS} \ chars.c \ mandoc.c \ mandoc_aux.c \ msec.c \ preconv.c \ read.c HTML_SRCS= eqn_html.c \ html.c \ man_html.c \ mdoc_html.c \ tbl_html.c MAN_SRCS= mdoc_man.c TERM_SRCS= eqn_term.c \ man_term.c \ mdoc_term.c \ term.c \ term_ascii.c \ term_ps.c \ tbl_term.c DB_SRCS= mandocdb.c \ mansearch.c \ mansearch_const.c \ manpath.c SRCS= ${LIB_SRCS} \ ${HTML_SRCS} \ ${MAN_SRCS} \ ${TERM_SRCS} \ main.c \ out.c \ tree.c SRCS+= ${DB_SRCS} WARNS?= 2 CFLAGS+= -DHAVE_CONFIG_H \ -I${.CURDIR}/../../lib/libohash/ \ -I${.CURDIR}/../../contrib/sqlite3 LIBADD= ohash sqlite3 .include Index: projects/clang-trunk/usr.sbin/pw/Makefile =================================================================== --- projects/clang-trunk/usr.sbin/pw/Makefile (revision 283870) +++ projects/clang-trunk/usr.sbin/pw/Makefile (revision 283871) @@ -1,19 +1,19 @@ # $FreeBSD$ PROG= pw MAN= pw.conf.5 pw.8 SRCS= pw.c pw_conf.c pw_user.c pw_group.c pw_log.c pw_nis.c pw_vpw.c \ grupd.c pwupd.c fileupd.c psdate.c \ bitmap.c cpdir.c rm_r.c -WARNS?= 2 +WARNS?= 3 LIBADD= crypt util sbuf .include .if ${MK_TESTS} != "no" SUBDIR+= tests .endif .include Index: projects/clang-trunk/usr.sbin/pw/grupd.c =================================================================== --- projects/clang-trunk/usr.sbin/pw/grupd.c (revision 283870) +++ projects/clang-trunk/usr.sbin/pw/grupd.c (revision 283871) @@ -1,124 +1,121 @@ /*- * Copyright (C) 1996 * David L. Nugent. All rights reserved. * * Redistribution and use in source and binary forms, with or without * modification, are permitted provided that the following conditions * are met: * 1. Redistributions of source code must retain the above copyright * notice, this list of conditions and the following disclaimer. * 2. Redistributions in binary form must reproduce the above copyright * notice, this list of conditions and the following disclaimer in the * documentation and/or other materials provided with the distribution. * * THIS SOFTWARE IS PROVIDED BY DAVID L. NUGENT AND CONTRIBUTORS ``AS IS'' AND * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE * ARE DISCLAIMED. IN NO EVENT SHALL DAVID L. NUGENT OR CONTRIBUTORS BE LIABLE * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF * SUCH DAMAGE. */ #ifndef lint static const char rcsid[] = "$FreeBSD$"; #endif /* not lint */ #include #include #include #include #include #include #include #include "pwupd.h" static char * grpath = _PATH_PWD; int setgrdir(const char * dir) { if (dir == NULL) return -1; else grpath = strdup(dir); if (grpath == NULL) return -1; return 0; } char * getgrpath(const char * file) { static char pathbuf[MAXPATHLEN]; snprintf(pathbuf, sizeof pathbuf, "%s/%s", grpath, file); return pathbuf; } static int gr_update(struct group * grp, char const * group) { int pfd, tfd; struct group *gr = NULL; struct group *old_gr = NULL; if (grp != NULL) gr = gr_dup(grp); if (group != NULL) old_gr = GETGRNAM(group); if (gr_init(grpath, NULL)) err(1, "gr_init()"); if ((pfd = gr_lock()) == -1) { gr_fini(); err(1, "gr_lock()"); } if ((tfd = gr_tmp(-1)) == -1) { gr_fini(); err(1, "gr_tmp()"); } if (gr_copy(pfd, tfd, gr, old_gr) == -1) { gr_fini(); err(1, "gr_copy()"); } if (gr_mkdb() == -1) { gr_fini(); err(1, "gr_mkdb()"); } free(gr); gr_fini(); return 0; } int addgrent(struct group * grp) { return gr_update(grp, NULL); } int chggrent(char const * login, struct group * grp) { return gr_update(grp, login); } int delgrent(struct group * grp) { - char group[MAXLOGNAME]; - strlcpy(group, grp->gr_name, MAXLOGNAME); - - return gr_update(NULL, group); + return (gr_update(NULL, grp->gr_name)); } Index: projects/clang-trunk/usr.sbin/pw/pw.c =================================================================== --- projects/clang-trunk/usr.sbin/pw/pw.c (revision 283870) +++ projects/clang-trunk/usr.sbin/pw/pw.c (revision 283871) @@ -1,460 +1,461 @@ /*- * Copyright (C) 1996 * David L. Nugent. All rights reserved. * * Redistribution and use in source and binary forms, with or without * modification, are permitted provided that the following conditions * are met: * 1. Redistributions of source code must retain the above copyright * notice, this list of conditions and the following disclaimer. * 2. Redistributions in binary form must reproduce the above copyright * notice, this list of conditions and the following disclaimer in the * documentation and/or other materials provided with the distribution. * * THIS SOFTWARE IS PROVIDED BY DAVID L. NUGENT AND CONTRIBUTORS ``AS IS'' AND * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE * ARE DISCLAIMED. IN NO EVENT SHALL DAVID L. NUGENT OR CONTRIBUTORS BE LIABLE * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF * SUCH DAMAGE. */ #ifndef lint static const char rcsid[] = "$FreeBSD$"; #endif /* not lint */ #include #include #include #include #include #include "pw.h" #if !defined(_PATH_YP) #define _PATH_YP "/var/yp/" #endif const char *Modes[] = { "add", "del", "mod", "show", "next", NULL}; const char *Which[] = {"user", "group", NULL}; static const char *Combo1[] = { "useradd", "userdel", "usermod", "usershow", "usernext", "lock", "unlock", "groupadd", "groupdel", "groupmod", "groupshow", "groupnext", NULL}; static const char *Combo2[] = { "adduser", "deluser", "moduser", "showuser", "nextuser", "lock", "unlock", "addgroup", "delgroup", "modgroup", "showgroup", "nextgroup", NULL}; struct pwf PWF = { 0, setpwent, endpwent, getpwent, getpwuid, getpwnam, setgrent, endgrent, getgrent, getgrgid, getgrnam, }; struct pwf VPWF = { 1, vsetpwent, vendpwent, vgetpwent, vgetpwuid, vgetpwnam, vsetgrent, vendgrent, vgetgrent, vgetgrgid, vgetgrnam, }; static struct cargs arglist; static int getindex(const char *words[], const char *word); static void cmdhelp(int mode, int which); int main(int argc, char *argv[]) { int ch; int mode = -1; int which = -1; char *config = NULL; struct userconf *cnf; struct stat st; static const char *opts[W_NUM][M_NUM] = { { /* user */ "V:C:qn:u:c:d:e:p:g:G:mM:k:s:oL:i:w:h:H:Db:NPy:Y", "V:C:qn:u:rY", "V:C:qn:u:c:d:e:p:g:G:mM:l:k:s:w:L:h:H:FNPY", "V:C:qn:u:FPa7", "V:C:q", "V:C:q", "V:C:q" }, { /* grp */ "V:C:qn:g:h:H:M:opNPY", "V:C:qn:g:Y", "V:C:qn:d:g:l:h:H:FM:m:NPY", "V:C:qn:g:FPa", "V:C:q" } }; static int (*funcs[W_NUM]) (struct userconf * _cnf, int _mode, struct cargs * _args) = { /* Request handlers */ pw_user, pw_group }; LIST_INIT(&arglist); (void)setlocale(LC_ALL, ""); /* * Break off the first couple of words to determine what exactly * we're being asked to do */ while (argc > 1) { int tmp; if (*argv[1] == '-') { /* * Special case, allow pw -V [args] for scripts etc. */ if (argv[1][1] == 'V') { optarg = &argv[1][2]; if (*optarg == '\0') { if (stat(argv[2], &st) != 0) errx(EX_OSFILE, \ "no such directory `%s'", argv[2]); if (!S_ISDIR(st.st_mode)) errx(EX_OSFILE, "`%s' not a " "directory", argv[2]); optarg = argv[2]; ++argv; --argc; } addarg(&arglist, 'V', optarg); } else break; } else if (mode == -1 && (tmp = getindex(Modes, argv[1])) != -1) mode = tmp; else if (which == -1 && (tmp = getindex(Which, argv[1])) != -1) which = tmp; else if ((mode == -1 && which == -1) && ((tmp = getindex(Combo1, argv[1])) != -1 || (tmp = getindex(Combo2, argv[1])) != -1)) { which = tmp / M_NUM; mode = tmp % M_NUM; } else if (strcmp(argv[1], "help") == 0 && argv[2] == NULL) cmdhelp(mode, which); else if (which != -1 && mode != -1) addarg(&arglist, 'n', argv[1]); else errx(EX_USAGE, "unknown keyword `%s'", argv[1]); ++argv; --argc; } /* * Bail out unless the user is specific! */ if (mode == -1 || which == -1) cmdhelp(mode, which); /* * We know which mode we're in and what we're about to do, so now * let's dispatch the remaining command line args in a genric way. */ optarg = NULL; while ((ch = getopt(argc, argv, opts[which][mode])) != -1) { if (ch == '?') errx(EX_USAGE, "unknown switch"); else addarg(&arglist, ch, optarg); optarg = NULL; } /* * Must be root to attempt an update */ if (geteuid() != 0 && mode != M_PRINT && mode != M_NEXT && getarg(&arglist, 'N')==NULL) errx(EX_NOPERM, "you must be root to run this program"); /* * We should immediately look for the -q 'quiet' switch so that we * don't bother with extraneous errors */ if (getarg(&arglist, 'q') != NULL) freopen(_PATH_DEVNULL, "w", stderr); /* * Set our base working path if not overridden */ config = getarg(&arglist, 'C') ? getarg(&arglist, 'C')->val : NULL; if (getarg(&arglist, 'V') != NULL) { char * etcpath = getarg(&arglist, 'V')->val; if (*etcpath) { if (config == NULL) { /* Only override config location if -C not specified */ - config = malloc(MAXPATHLEN); - snprintf(config, MAXPATHLEN, "%s/pw.conf", etcpath); + asprintf(&config, "%s/pw.conf", etcpath); + if (config == NULL) + errx(EX_OSERR, "out of memory"); } memcpy(&PWF, &VPWF, sizeof PWF); setpwdir(etcpath); setgrdir(etcpath); } } /* * Now, let's do the common initialisation */ cnf = read_userconfig(config); ch = funcs[which] (cnf, mode, &arglist); /* * If everything went ok, and we've been asked to update * the NIS maps, then do it now */ if (ch == EXIT_SUCCESS && getarg(&arglist, 'Y') != NULL) { pid_t pid; fflush(NULL); if (chdir(_PATH_YP) == -1) warn("chdir(" _PATH_YP ")"); else if ((pid = fork()) == -1) warn("fork()"); else if (pid == 0) { /* Is make anywhere else? */ execlp("/usr/bin/make", "make", (char *)NULL); _exit(1); } else { int i; waitpid(pid, &i, 0); if ((i = WEXITSTATUS(i)) != 0) errx(ch, "make exited with status %d", i); else pw_log(cnf, mode, which, "NIS maps updated"); } } return ch; } static int getindex(const char *words[], const char *word) { int i = 0; while (words[i]) { if (strcmp(words[i], word) == 0) return i; i++; } return -1; } /* * This is probably an overkill for a cmdline help system, but it reflects * the complexity of the command line. */ static void cmdhelp(int mode, int which) { if (which == -1) fprintf(stderr, "usage:\n pw [user|group|lock|unlock] [add|del|mod|show|next] [help|switches/values]\n"); else if (mode == -1) fprintf(stderr, "usage:\n pw %s [add|del|mod|show|next] [help|switches/values]\n", Which[which]); else { /* * We need to give mode specific help */ static const char *help[W_NUM][M_NUM] = { { "usage: pw useradd [name] [switches]\n" "\t-V etcdir alternate /etc location\n" "\t-C config configuration file\n" "\t-q quiet operation\n" " Adding users:\n" "\t-n name login name\n" "\t-u uid user id\n" "\t-c comment user name/comment\n" "\t-d directory home directory\n" "\t-e date account expiry date\n" "\t-p date password expiry date\n" "\t-g grp initial group\n" "\t-G grp1,grp2 additional groups\n" "\t-m [ -k dir ] create and set up home\n" "\t-M mode home directory permissions\n" "\t-s shell name of login shell\n" "\t-o duplicate uid ok\n" "\t-L class user class\n" "\t-h fd read password on fd\n" "\t-H fd read encrypted password on fd\n" "\t-Y update NIS maps\n" "\t-N no update\n" " Setting defaults:\n" "\t-V etcdir alternate /etc location\n" "\t-D set user defaults\n" "\t-b dir default home root dir\n" "\t-e period default expiry period\n" "\t-p period default password change period\n" "\t-g group default group\n" "\t-G grp1,grp2 additional groups\n" "\t-L class default user class\n" "\t-k dir default home skeleton\n" "\t-M mode home directory permissions\n" "\t-u min,max set min,max uids\n" "\t-i min,max set min,max gids\n" "\t-w method set default password method\n" "\t-s shell default shell\n" "\t-y path set NIS passwd file path\n", "usage: pw userdel [uid|name] [switches]\n" "\t-V etcdir alternate /etc location\n" "\t-n name login name\n" "\t-u uid user id\n" "\t-Y update NIS maps\n" "\t-r remove home & contents\n", "usage: pw usermod [uid|name] [switches]\n" "\t-V etcdir alternate /etc location\n" "\t-C config configuration file\n" "\t-q quiet operation\n" "\t-F force add if no user\n" "\t-n name login name\n" "\t-u uid user id\n" "\t-c comment user name/comment\n" "\t-d directory home directory\n" "\t-e date account expiry date\n" "\t-p date password expiry date\n" "\t-g grp initial group\n" "\t-G grp1,grp2 additional groups\n" "\t-l name new login name\n" "\t-L class user class\n" "\t-m [ -k dir ] create and set up home\n" "\t-M mode home directory permissions\n" "\t-s shell name of login shell\n" "\t-w method set new password using method\n" "\t-h fd read password on fd\n" "\t-H fd read encrypted password on fd\n" "\t-Y update NIS maps\n" "\t-N no update\n", "usage: pw usershow [uid|name] [switches]\n" "\t-V etcdir alternate /etc location\n" "\t-n name login name\n" "\t-u uid user id\n" "\t-F force print\n" "\t-P prettier format\n" "\t-a print all users\n" "\t-7 print in v7 format\n", "usage: pw usernext [switches]\n" "\t-V etcdir alternate /etc location\n" "\t-C config configuration file\n" "\t-q quiet operation\n", "usage pw: lock [switches]\n" "\t-V etcdir alternate /etc locations\n" "\t-C config configuration file\n" "\t-q quiet operation\n", "usage pw: unlock [switches]\n" "\t-V etcdir alternate /etc locations\n" "\t-C config configuration file\n" "\t-q quiet operation\n" }, { "usage: pw groupadd [group|gid] [switches]\n" "\t-V etcdir alternate /etc location\n" "\t-C config configuration file\n" "\t-q quiet operation\n" "\t-n group group name\n" "\t-g gid group id\n" "\t-M usr1,usr2 add users as group members\n" "\t-o duplicate gid ok\n" "\t-Y update NIS maps\n" "\t-N no update\n", "usage: pw groupdel [group|gid] [switches]\n" "\t-V etcdir alternate /etc location\n" "\t-n name group name\n" "\t-g gid group id\n" "\t-Y update NIS maps\n", "usage: pw groupmod [group|gid] [switches]\n" "\t-V etcdir alternate /etc location\n" "\t-C config configuration file\n" "\t-q quiet operation\n" "\t-F force add if not exists\n" "\t-n name group name\n" "\t-g gid group id\n" "\t-M usr1,usr2 replaces users as group members\n" "\t-m usr1,usr2 add users as group members\n" "\t-d usr1,usr2 delete users as group members\n" "\t-l name new group name\n" "\t-Y update NIS maps\n" "\t-N no update\n", "usage: pw groupshow [group|gid] [switches]\n" "\t-V etcdir alternate /etc location\n" "\t-n name group name\n" "\t-g gid group id\n" "\t-F force print\n" "\t-P prettier format\n" "\t-a print all accounting groups\n", "usage: pw groupnext [switches]\n" "\t-V etcdir alternate /etc location\n" "\t-C config configuration file\n" "\t-q quiet operation\n" } }; fprintf(stderr, "%s", help[which][mode]); } exit(EXIT_FAILURE); } struct carg * getarg(struct cargs * _args, int ch) { struct carg *c = LIST_FIRST(_args); while (c != NULL && c->ch != ch) c = LIST_NEXT(c, list); return c; } struct carg * addarg(struct cargs * _args, int ch, char *argstr) { struct carg *ca = malloc(sizeof(struct carg)); if (ca == NULL) errx(EX_OSERR, "out of memory"); ca->ch = ch; ca->val = argstr; LIST_INSERT_HEAD(_args, ca, list); return ca; } Index: projects/clang-trunk/usr.sbin/pw/pw_conf.c =================================================================== --- projects/clang-trunk/usr.sbin/pw/pw_conf.c (revision 283870) +++ projects/clang-trunk/usr.sbin/pw/pw_conf.c (revision 283871) @@ -1,499 +1,500 @@ /*- * Copyright (C) 1996 * David L. Nugent. All rights reserved. * * Redistribution and use in source and binary forms, with or without * modification, are permitted provided that the following conditions * are met: * 1. Redistributions of source code must retain the above copyright * notice, this list of conditions and the following disclaimer. * 2. Redistributions in binary form must reproduce the above copyright * notice, this list of conditions and the following disclaimer in the * documentation and/or other materials provided with the distribution. * * THIS SOFTWARE IS PROVIDED BY DAVID L. NUGENT AND CONTRIBUTORS ``AS IS'' AND * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE * ARE DISCLAIMED. IN NO EVENT SHALL DAVID L. NUGENT OR CONTRIBUTORS BE LIABLE * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF * SUCH DAMAGE. */ #ifndef lint static const char rcsid[] = "$FreeBSD$"; #endif /* not lint */ #include #include #include #include #include #include #include "pw.h" #define debugging 0 enum { _UC_NONE, _UC_DEFAULTPWD, _UC_REUSEUID, _UC_REUSEGID, _UC_NISPASSWD, _UC_DOTDIR, _UC_NEWMAIL, _UC_LOGFILE, _UC_HOMEROOT, _UC_HOMEMODE, _UC_SHELLPATH, _UC_SHELLS, _UC_DEFAULTSHELL, _UC_DEFAULTGROUP, _UC_EXTRAGROUPS, _UC_DEFAULTCLASS, _UC_MINUID, _UC_MAXUID, _UC_MINGID, _UC_MAXGID, _UC_EXPIRE, _UC_PASSWORD, _UC_FIELDS }; static char bourne_shell[] = "sh"; static char *system_shells[_UC_MAXSHELLS] = { bourne_shell, "csh", "tcsh" }; static char const *booltrue[] = { "yes", "true", "1", "on", NULL }; static char const *boolfalse[] = { "no", "false", "0", "off", NULL }; static struct userconf config = { 0, /* Default password for new users? (nologin) */ 0, /* Reuse uids? */ 0, /* Reuse gids? */ NULL, /* NIS version of the passwd file */ "/usr/share/skel", /* Where to obtain skeleton files */ NULL, /* Mail to send to new accounts */ "/var/log/userlog", /* Where to log changes */ "/home", /* Where to create home directory */ _DEF_DIRMODE, /* Home directory perms, modified by umask */ "/bin", /* Where shells are located */ system_shells, /* List of shells (first is default) */ bourne_shell, /* Default shell */ NULL, /* Default group name */ NULL, /* Default (additional) groups */ NULL, /* Default login class */ 1000, 32000, /* Allowed range of uids */ 1000, 32000, /* Allowed range of gids */ 0, /* Days until account expires */ 0, /* Days until password expires */ 0 /* size of default_group array */ }; static char const *comments[_UC_FIELDS] = { "#\n# pw.conf - user/group configuration defaults\n#\n", "\n# Password for new users? no=nologin yes=loginid none=blank random=random\n", "\n# Reuse gaps in uid sequence? (yes or no)\n", "\n# Reuse gaps in gid sequence? (yes or no)\n", "\n# Path to the NIS passwd file (blank or 'no' for none)\n", "\n# Obtain default dotfiles from this directory\n", "\n# Mail this file to new user (/etc/newuser.msg or no)\n", "\n# Log add/change/remove information in this file\n", "\n# Root directory in which $HOME directory is created\n", "\n# Mode for the new $HOME directory, will be modified by umask\n", "\n# Colon separated list of directories containing valid shells\n", "\n# Comma separated list of available shells (without paths)\n", "\n# Default shell (without path)\n", "\n# Default group (leave blank for new group per user)\n", "\n# Extra groups for new users\n", "\n# Default login class for new users\n", "\n# Range of valid default user ids\n", NULL, "\n# Range of valid default group ids\n", NULL, "\n# Days after which account expires (0=disabled)\n", "\n# Days after which password expires (0=disabled)\n" }; static char const *kwds[] = { "", "defaultpasswd", "reuseuids", "reusegids", "nispasswd", "skeleton", "newmail", "logfile", "home", "homemode", "shellpath", "shells", "defaultshell", "defaultgroup", "extragroups", "defaultclass", "minuid", "maxuid", "mingid", "maxgid", "expire_days", "password_days", NULL }; static char * unquote(char const * str) { if (str && (*str == '"' || *str == '\'')) { char *p = strchr(str + 1, *str); if (p != NULL) *p = '\0'; return (char *) (*++str ? str : NULL); } return (char *) str; } int boolean_val(char const * str, int dflt) { if ((str = unquote(str)) != NULL) { int i; for (i = 0; booltrue[i]; i++) if (strcmp(str, booltrue[i]) == 0) return 1; for (i = 0; boolfalse[i]; i++) if (strcmp(str, boolfalse[i]) == 0) return 0; /* * Special cases for defaultpassword */ if (strcmp(str, "random") == 0) return -1; if (strcmp(str, "none") == 0) return -2; } return dflt; } char const * boolean_str(int val) { if (val == -1) return "random"; else if (val == -2) return "none"; else return val ? booltrue[0] : boolfalse[0]; } char * newstr(char const * p) { char *q; if ((p = unquote(p)) == NULL) return (NULL); if ((q = strdup(p)) == NULL) err(1, "strdup()"); return (q); } struct userconf * read_userconfig(char const * file) { FILE *fp; char *buf, *p; size_t linecap; ssize_t linelen; buf = NULL; linecap = 0; config.numgroups = 200; config.groups = calloc(config.numgroups, sizeof(char *)); if (config.groups == NULL) err(1, "calloc()"); if (file == NULL) file = _PATH_PW_CONF; - if ((fp = fopen(file, "r")) != NULL) { - while ((linelen = getline(&buf, &linecap, fp)) > 0) { - if (*buf && (p = strtok(buf, " \t\r\n=")) != NULL && *p != '#') { - static char const toks[] = " \t\r\n,="; - char *q = strtok(NULL, toks); - int i = 0; - mode_t *modeset; + if ((fp = fopen(file, "r")) == NULL) + return (&config); - while (i < _UC_FIELDS && strcmp(p, kwds[i]) != 0) - ++i; + while ((linelen = getline(&buf, &linecap, fp)) > 0) { + if (*buf && (p = strtok(buf, " \t\r\n=")) != NULL && *p != '#') { + static char const toks[] = " \t\r\n,="; + char *q = strtok(NULL, toks); + int i = 0; + mode_t *modeset; + + while (i < _UC_FIELDS && strcmp(p, kwds[i]) != 0) + ++i; #if debugging - if (i == _UC_FIELDS) - printf("Got unknown kwd `%s' val=`%s'\n", p, q ? q : ""); - else - printf("Got kwd[%s]=%s\n", p, q); + if (i == _UC_FIELDS) + printf("Got unknown kwd `%s' val=`%s'\n", p, q ? q : ""); + else + printf("Got kwd[%s]=%s\n", p, q); #endif - switch (i) { - case _UC_DEFAULTPWD: - config.default_password = boolean_val(q, 1); - break; - case _UC_REUSEUID: - config.reuse_uids = boolean_val(q, 0); - break; - case _UC_REUSEGID: - config.reuse_gids = boolean_val(q, 0); - break; - case _UC_NISPASSWD: - config.nispasswd = (q == NULL || !boolean_val(q, 1)) - ? NULL : newstr(q); - break; - case _UC_DOTDIR: - config.dotdir = (q == NULL || !boolean_val(q, 1)) - ? NULL : newstr(q); - break; + switch (i) { + case _UC_DEFAULTPWD: + config.default_password = boolean_val(q, 1); + break; + case _UC_REUSEUID: + config.reuse_uids = boolean_val(q, 0); + break; + case _UC_REUSEGID: + config.reuse_gids = boolean_val(q, 0); + break; + case _UC_NISPASSWD: + config.nispasswd = (q == NULL || !boolean_val(q, 1)) + ? NULL : newstr(q); + break; + case _UC_DOTDIR: + config.dotdir = (q == NULL || !boolean_val(q, 1)) + ? NULL : newstr(q); + break; case _UC_NEWMAIL: - config.newmail = (q == NULL || !boolean_val(q, 1)) - ? NULL : newstr(q); - break; - case _UC_LOGFILE: - config.logfile = (q == NULL || !boolean_val(q, 1)) - ? NULL : newstr(q); - break; - case _UC_HOMEROOT: - config.home = (q == NULL || !boolean_val(q, 1)) - ? "/home" : newstr(q); - break; - case _UC_HOMEMODE: - modeset = setmode(q); - config.homemode = (q == NULL || !boolean_val(q, 1)) - ? _DEF_DIRMODE : getmode(modeset, _DEF_DIRMODE); - free(modeset); - break; - case _UC_SHELLPATH: - config.shelldir = (q == NULL || !boolean_val(q, 1)) - ? "/bin" : newstr(q); - break; - case _UC_SHELLS: - for (i = 0; i < _UC_MAXSHELLS && q != NULL; i++, q = strtok(NULL, toks)) - system_shells[i] = newstr(q); - if (i > 0) - while (i < _UC_MAXSHELLS) - system_shells[i++] = NULL; - break; - case _UC_DEFAULTSHELL: - config.shell_default = (q == NULL || !boolean_val(q, 1)) - ? (char *) bourne_shell : newstr(q); - break; - case _UC_DEFAULTGROUP: - q = unquote(q); - config.default_group = (q == NULL || !boolean_val(q, 1) || GETGRNAM(q) == NULL) - ? NULL : newstr(q); - break; - case _UC_EXTRAGROUPS: - for (i = 0; q != NULL; q = strtok(NULL, toks)) { - if (extendarray(&config.groups, &config.numgroups, i + 2) != -1) - config.groups[i++] = newstr(q); - } - if (i > 0) - while (i < config.numgroups) - config.groups[i++] = NULL; - break; - case _UC_DEFAULTCLASS: - config.default_class = (q == NULL || !boolean_val(q, 1)) - ? NULL : newstr(q); - break; - case _UC_MINUID: - if ((q = unquote(q)) != NULL && isdigit(*q)) - config.min_uid = (uid_t) atol(q); - break; - case _UC_MAXUID: - if ((q = unquote(q)) != NULL && isdigit(*q)) - config.max_uid = (uid_t) atol(q); - break; - case _UC_MINGID: - if ((q = unquote(q)) != NULL && isdigit(*q)) - config.min_gid = (gid_t) atol(q); - break; - case _UC_MAXGID: - if ((q = unquote(q)) != NULL && isdigit(*q)) - config.max_gid = (gid_t) atol(q); - break; - case _UC_EXPIRE: - if ((q = unquote(q)) != NULL && isdigit(*q)) - config.expire_days = atoi(q); - break; - case _UC_PASSWORD: - if ((q = unquote(q)) != NULL && isdigit(*q)) - config.password_days = atoi(q); - break; - case _UC_FIELDS: - case _UC_NONE: - break; + config.newmail = (q == NULL || !boolean_val(q, 1)) + ? NULL : newstr(q); + break; + case _UC_LOGFILE: + config.logfile = (q == NULL || !boolean_val(q, 1)) + ? NULL : newstr(q); + break; + case _UC_HOMEROOT: + config.home = (q == NULL || !boolean_val(q, 1)) + ? "/home" : newstr(q); + break; + case _UC_HOMEMODE: + modeset = setmode(q); + config.homemode = (q == NULL || !boolean_val(q, 1)) + ? _DEF_DIRMODE : getmode(modeset, _DEF_DIRMODE); + free(modeset); + break; + case _UC_SHELLPATH: + config.shelldir = (q == NULL || !boolean_val(q, 1)) + ? "/bin" : newstr(q); + break; + case _UC_SHELLS: + for (i = 0; i < _UC_MAXSHELLS && q != NULL; i++, q = strtok(NULL, toks)) + system_shells[i] = newstr(q); + if (i > 0) + while (i < _UC_MAXSHELLS) + system_shells[i++] = NULL; + break; + case _UC_DEFAULTSHELL: + config.shell_default = (q == NULL || !boolean_val(q, 1)) + ? (char *) bourne_shell : newstr(q); + break; + case _UC_DEFAULTGROUP: + q = unquote(q); + config.default_group = (q == NULL || !boolean_val(q, 1) || GETGRNAM(q) == NULL) + ? NULL : newstr(q); + break; + case _UC_EXTRAGROUPS: + for (i = 0; q != NULL; q = strtok(NULL, toks)) { + if (extendarray(&config.groups, &config.numgroups, i + 2) != -1) + config.groups[i++] = newstr(q); } + if (i > 0) + while (i < config.numgroups) + config.groups[i++] = NULL; + break; + case _UC_DEFAULTCLASS: + config.default_class = (q == NULL || !boolean_val(q, 1)) + ? NULL : newstr(q); + break; + case _UC_MINUID: + if ((q = unquote(q)) != NULL && isdigit(*q)) + config.min_uid = (uid_t) atol(q); + break; + case _UC_MAXUID: + if ((q = unquote(q)) != NULL && isdigit(*q)) + config.max_uid = (uid_t) atol(q); + break; + case _UC_MINGID: + if ((q = unquote(q)) != NULL && isdigit(*q)) + config.min_gid = (gid_t) atol(q); + break; + case _UC_MAXGID: + if ((q = unquote(q)) != NULL && isdigit(*q)) + config.max_gid = (gid_t) atol(q); + break; + case _UC_EXPIRE: + if ((q = unquote(q)) != NULL && isdigit(*q)) + config.expire_days = atoi(q); + break; + case _UC_PASSWORD: + if ((q = unquote(q)) != NULL && isdigit(*q)) + config.password_days = atoi(q); + break; + case _UC_FIELDS: + case _UC_NONE: + break; } } - if (linecap > 0) - free(buf); - fclose(fp); } - return &config; + free(buf); + fclose(fp); + + return (&config); } int write_userconfig(char const * file) { int fd; int i, j; struct sbuf *buf; FILE *fp; if (file == NULL) file = _PATH_PW_CONF; if ((fd = open(file, O_CREAT|O_RDWR|O_TRUNC|O_EXLOCK, 0644)) == -1) return (0); if ((fp = fdopen(fd, "w")) == NULL) { close(fd); return (0); } buf = sbuf_new_auto(); for (i = _UC_NONE; i < _UC_FIELDS; i++) { int quote = 1; sbuf_clear(buf); switch (i) { case _UC_DEFAULTPWD: sbuf_cat(buf, boolean_str(config.default_password)); break; case _UC_REUSEUID: sbuf_cat(buf, boolean_str(config.reuse_uids)); break; case _UC_REUSEGID: sbuf_cat(buf, boolean_str(config.reuse_gids)); break; case _UC_NISPASSWD: sbuf_cat(buf, config.nispasswd ? config.nispasswd : ""); quote = 0; break; case _UC_DOTDIR: sbuf_cat(buf, config.dotdir ? config.dotdir : boolean_str(0)); break; case _UC_NEWMAIL: sbuf_cat(buf, config.newmail ? config.newmail : boolean_str(0)); break; case _UC_LOGFILE: sbuf_cat(buf, config.logfile ? config.logfile : boolean_str(0)); break; case _UC_HOMEROOT: sbuf_cat(buf, config.home); break; case _UC_HOMEMODE: sbuf_printf(buf, "%04o", config.homemode); quote = 0; break; case _UC_SHELLPATH: sbuf_cat(buf, config.shelldir); break; case _UC_SHELLS: for (j = 0; j < _UC_MAXSHELLS && system_shells[j] != NULL; j++) sbuf_printf(buf, "%s\"%s\"", j ? "," : "", system_shells[j]); quote = 0; break; case _UC_DEFAULTSHELL: sbuf_cat(buf, config.shell_default ? config.shell_default : bourne_shell); break; case _UC_DEFAULTGROUP: sbuf_cat(buf, config.default_group ? config.default_group : ""); break; case _UC_EXTRAGROUPS: for (j = 0; j < config.numgroups && config.groups[j] != NULL; j++) sbuf_printf(buf, "%s\"%s\"", j ? "," : "", config.groups[j]); quote = 0; break; case _UC_DEFAULTCLASS: sbuf_cat(buf, config.default_class ? config.default_class : ""); break; case _UC_MINUID: - sbuf_printf(buf, "%lu", (unsigned long) config.min_uid); + sbuf_printf(buf, "%u", config.min_uid); quote = 0; break; case _UC_MAXUID: - sbuf_printf(buf, "%lu", (unsigned long) config.max_uid); + sbuf_printf(buf, "%u", config.max_uid); quote = 0; break; case _UC_MINGID: - sbuf_printf(buf, "%lu", (unsigned long) config.min_gid); + sbuf_printf(buf, "%u", config.min_gid); quote = 0; break; case _UC_MAXGID: - sbuf_printf(buf, "%lu", (unsigned long) config.max_gid); + sbuf_printf(buf, "%u", config.max_gid); quote = 0; break; case _UC_EXPIRE: sbuf_printf(buf, "%d", config.expire_days); quote = 0; break; case _UC_PASSWORD: sbuf_printf(buf, "%d", config.password_days); quote = 0; break; case _UC_NONE: break; } sbuf_finish(buf); if (comments[i]) fputs(comments[i], fp); if (*kwds[i]) { if (quote) fprintf(fp, "%s = \"%s\"\n", kwds[i], sbuf_data(buf)); else fprintf(fp, "%s = %s\n", kwds[i], sbuf_data(buf)); #if debugging printf("WROTE: %s = %s\n", kwds[i], sbuf_data(buf)); #endif } } sbuf_delete(buf); return (fclose(fp) != EOF); } Index: projects/clang-trunk/usr.sbin/pw/pw_group.c =================================================================== --- projects/clang-trunk/usr.sbin/pw/pw_group.c (revision 283870) +++ projects/clang-trunk/usr.sbin/pw/pw_group.c (revision 283871) @@ -1,438 +1,435 @@ /*- * Copyright (C) 1996 * David L. Nugent. All rights reserved. * * Redistribution and use in source and binary forms, with or without * modification, are permitted provided that the following conditions * are met: * 1. Redistributions of source code must retain the above copyright * notice, this list of conditions and the following disclaimer. * 2. Redistributions in binary form must reproduce the above copyright * notice, this list of conditions and the following disclaimer in the * documentation and/or other materials provided with the distribution. * * THIS SOFTWARE IS PROVIDED BY DAVID L. NUGENT AND CONTRIBUTORS ``AS IS'' AND * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE * ARE DISCLAIMED. IN NO EVENT SHALL DAVID L. NUGENT OR CONTRIBUTORS BE LIABLE * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF * SUCH DAMAGE. */ #ifndef lint static const char rcsid[] = "$FreeBSD$"; #endif /* not lint */ #include #include #include #include #include #include #include #include "pw.h" #include "bitmap.h" static struct passwd *lookup_pwent(const char *user); static void delete_members(char ***members, int *grmembers, int *i, struct carg *arg, struct group *grp); static int print_group(struct group * grp, int pretty); static gid_t gr_gidpolicy(struct userconf * cnf, struct cargs * args); int pw_group(struct userconf * cnf, int mode, struct cargs * args) { int rc; struct carg *a_newname = getarg(args, 'l'); struct carg *a_name = getarg(args, 'n'); struct carg *a_gid = getarg(args, 'g'); struct carg *arg; struct group *grp = NULL; int grmembers = 0; char **members = NULL; static struct group fakegroup = { "nogroup", "*", -1, NULL }; if (a_gid != NULL) { if (strspn(a_gid->val, "0123456789") != strlen(a_gid->val)) errx(EX_USAGE, "-g expects a number"); } if (mode == M_LOCK || mode == M_UNLOCK) errx(EX_USAGE, "'lock' command is not available for groups"); /* * With M_NEXT, we only need to return the * next gid to stdout */ if (mode == M_NEXT) { gid_t next = gr_gidpolicy(cnf, args); if (getarg(args, 'q')) return next; - printf("%ld\n", (long)next); + printf("%u\n", next); return EXIT_SUCCESS; } if (mode == M_PRINT && getarg(args, 'a')) { int pretty = getarg(args, 'P') != NULL; SETGRENT(); while ((grp = GETGRENT()) != NULL) print_group(grp, pretty); ENDGRENT(); return EXIT_SUCCESS; } if (a_gid == NULL) { if (a_name == NULL) errx(EX_DATAERR, "group name or id required"); if (mode != M_ADD && grp == NULL && isdigit((unsigned char)*a_name->val)) { (a_gid = a_name)->ch = 'g'; a_name = NULL; } } grp = (a_name != NULL) ? GETGRNAM(a_name->val) : GETGRGID((gid_t) atoi(a_gid->val)); if (mode == M_UPDATE || mode == M_DELETE || mode == M_PRINT) { if (a_name == NULL && grp == NULL) /* Try harder */ grp = GETGRGID(atoi(a_gid->val)); if (grp == NULL) { if (mode == M_PRINT && getarg(args, 'F')) { char *fmems[1]; fmems[0] = NULL; fakegroup.gr_name = a_name ? a_name->val : "nogroup"; - fakegroup.gr_gid = a_gid ? (gid_t) atol(a_gid->val) : -1; + fakegroup.gr_gid = a_gid ? (gid_t) atol(a_gid->val) : (gid_t)-1; fakegroup.gr_mem = fmems; return print_group(&fakegroup, getarg(args, 'P') != NULL); } errx(EX_DATAERR, "unknown group `%s'", a_name ? a_name->val : a_gid->val); } if (a_name == NULL) /* Needed later */ a_name = addarg(args, 'n', grp->gr_name); /* * Handle deletions now */ if (mode == M_DELETE) { gid_t gid = grp->gr_gid; rc = delgrent(grp); if (rc == -1) err(EX_IOERR, "group '%s' not available (NIS?)", grp->gr_name); else if (rc != 0) { - warn("group update"); - return EX_IOERR; + err(EX_IOERR, "group update"); } - pw_log(cnf, mode, W_GROUP, "%s(%ld) removed", a_name->val, (long) gid); + pw_log(cnf, mode, W_GROUP, "%s(%u) removed", a_name->val, gid); return EXIT_SUCCESS; } else if (mode == M_PRINT) return print_group(grp, getarg(args, 'P') != NULL); if (a_gid) grp->gr_gid = (gid_t) atoi(a_gid->val); if (a_newname != NULL) grp->gr_name = pw_checkname((u_char *)a_newname->val, 0); } else { if (a_name == NULL) /* Required */ errx(EX_DATAERR, "group name required"); else if (grp != NULL) /* Exists */ errx(EX_DATAERR, "group name `%s' already exists", a_name->val); extendarray(&members, &grmembers, 200); members[0] = NULL; grp = &fakegroup; grp->gr_name = pw_checkname((u_char *)a_name->val, 0); grp->gr_passwd = "*"; grp->gr_gid = gr_gidpolicy(cnf, args); grp->gr_mem = members; } /* * This allows us to set a group password Group passwords is an * antique idea, rarely used and insecure (no secure database) Should * be discouraged, but it is apparently still supported by some * software. */ if ((arg = getarg(args, 'h')) != NULL || (arg = getarg(args, 'H')) != NULL) { if (strcmp(arg->val, "-") == 0) grp->gr_passwd = "*"; /* No access */ else { int fd = atoi(arg->val); int precrypt = (arg->ch == 'H'); int b; int istty = isatty(fd); struct termios t; char *p, line[256]; if (istty) { if (tcgetattr(fd, &t) == -1) istty = 0; else { struct termios n = t; /* Disable echo */ n.c_lflag &= ~(ECHO); tcsetattr(fd, TCSANOW, &n); printf("%sassword for group %s:", (mode == M_UPDATE) ? "New p" : "P", grp->gr_name); fflush(stdout); } } b = read(fd, line, sizeof(line) - 1); if (istty) { /* Restore state */ tcsetattr(fd, TCSANOW, &t); fputc('\n', stdout); fflush(stdout); } - if (b < 0) { - warn("-h file descriptor"); - return EX_OSERR; - } + if (b < 0) + err(EX_OSERR, "-h file descriptor"); line[b] = '\0'; if ((p = strpbrk(line, " \t\r\n")) != NULL) *p = '\0'; if (!*line) errx(EX_DATAERR, "empty password read on file descriptor %d", fd); if (precrypt) { if (strchr(line, ':') != NULL) return EX_DATAERR; grp->gr_passwd = line; } else grp->gr_passwd = pw_pwcrypt(line); } } if (((arg = getarg(args, 'M')) != NULL || (arg = getarg(args, 'd')) != NULL || (arg = getarg(args, 'm')) != NULL) && arg->val) { int i = 0; char *p; struct passwd *pwd; /* Make sure this is not stay NULL with -M "" */ extendarray(&members, &grmembers, 200); if (arg->ch == 'd') delete_members(&members, &grmembers, &i, arg, grp); else if (arg->ch == 'm') { int k = 0; if (grp->gr_mem != NULL) { while (grp->gr_mem[k] != NULL) { if (extendarray(&members, &grmembers, i + 2) != -1) members[i++] = grp->gr_mem[k]; k++; } } } if (arg->ch != 'd') for (p = strtok(arg->val, ", \t"); p != NULL; p = strtok(NULL, ", \t")) { int j; /* * Check for duplicates */ pwd = lookup_pwent(p); for (j = 0; j < i && strcmp(members[j], pwd->pw_name) != 0; j++) ; if (j == i && extendarray(&members, &grmembers, i + 2) != -1) members[i++] = newstr(pwd->pw_name); } while (i < grmembers) members[i++] = NULL; grp->gr_mem = members; } if (getarg(args, 'N') != NULL) return print_group(grp, getarg(args, 'P') != NULL); if (mode == M_ADD && (rc = addgrent(grp)) != 0) { if (rc == -1) - warnx("group '%s' already exists", grp->gr_name); + errx(EX_IOERR, "group '%s' already exists", + grp->gr_name); else - warn("group update"); - return EX_IOERR; + err(EX_IOERR, "group update"); } else if (mode == M_UPDATE && (rc = chggrent(a_name->val, grp)) != 0) { if (rc == -1) - warnx("group '%s' not available (NIS?)", grp->gr_name); + errx(EX_IOERR, "group '%s' not available (NIS?)", + grp->gr_name); else - warn("group update"); - return EX_IOERR; + err(EX_IOERR, "group update"); } arg = a_newname != NULL ? a_newname : a_name; /* grp may have been invalidated */ if ((grp = GETGRNAM(arg->val)) == NULL) errx(EX_SOFTWARE, "group disappeared during update"); - pw_log(cnf, mode, W_GROUP, "%s(%ld)", grp->gr_name, (long) grp->gr_gid); + pw_log(cnf, mode, W_GROUP, "%s(%u)", grp->gr_name, grp->gr_gid); free(members); return EXIT_SUCCESS; } /* * Lookup a passwd entry using a name or UID. */ static struct passwd * lookup_pwent(const char *user) { struct passwd *pwd; if ((pwd = GETPWNAM(user)) == NULL && (!isdigit((unsigned char)*user) || (pwd = getpwuid((uid_t) atoi(user))) == NULL)) errx(EX_NOUSER, "user `%s' does not exist", user); return (pwd); } /* * Delete requested members from a group. */ static void delete_members(char ***members, int *grmembers, int *i, struct carg *arg, struct group *grp) { bool matchFound; char *user; char *valueCopy; char *valuePtr; int k; struct passwd *pwd; if (grp->gr_mem == NULL) return; k = 0; while (grp->gr_mem[k] != NULL) { matchFound = false; if ((valueCopy = strdup(arg->val)) == NULL) errx(EX_UNAVAILABLE, "out of memory"); valuePtr = valueCopy; while ((user = strsep(&valuePtr, ", \t")) != NULL) { pwd = lookup_pwent(user); if (strcmp(grp->gr_mem[k], pwd->pw_name) == 0) { matchFound = true; break; } } free(valueCopy); if (!matchFound && extendarray(members, grmembers, *i + 2) != -1) (*members)[(*i)++] = grp->gr_mem[k]; k++; } return; } static gid_t gr_gidpolicy(struct userconf * cnf, struct cargs * args) { struct group *grp; gid_t gid = (gid_t) - 1; struct carg *a_gid = getarg(args, 'g'); /* * Check the given gid, if any */ if (a_gid != NULL) { gid = (gid_t) atol(a_gid->val); if ((grp = GETGRGID(gid)) != NULL && getarg(args, 'o') == NULL) - errx(EX_DATAERR, "gid `%ld' has already been allocated", (long) grp->gr_gid); + errx(EX_DATAERR, "gid `%u' has already been allocated", grp->gr_gid); } else { struct bitmap bm; /* * We need to allocate the next available gid under one of * two policies a) Grab the first unused gid b) Grab the * highest possible unused gid */ if (cnf->min_gid >= cnf->max_gid) { /* Sanity claus^H^H^H^Hheck */ cnf->min_gid = 1000; cnf->max_gid = 32000; } bm = bm_alloc(cnf->max_gid - cnf->min_gid + 1); /* * Now, let's fill the bitmap from the password file */ SETGRENT(); while ((grp = GETGRENT()) != NULL) if ((gid_t)grp->gr_gid >= (gid_t)cnf->min_gid && (gid_t)grp->gr_gid <= (gid_t)cnf->max_gid) bm_setbit(&bm, grp->gr_gid - cnf->min_gid); ENDGRENT(); /* * Then apply the policy, with fallback to reuse if necessary */ if (cnf->reuse_gids) gid = (gid_t) (bm_firstunset(&bm) + cnf->min_gid); else { gid = (gid_t) (bm_lastset(&bm) + 1); if (!bm_isset(&bm, gid)) gid += cnf->min_gid; else gid = (gid_t) (bm_firstunset(&bm) + cnf->min_gid); } /* * Another sanity check */ if (gid < cnf->min_gid || gid > cnf->max_gid) errx(EX_SOFTWARE, "unable to allocate a new gid - range fully used"); bm_dealloc(&bm); } return gid; } static int print_group(struct group * grp, int pretty) { if (!pretty) { char *buf = NULL; buf = gr_make(grp); printf("%s\n", buf); free(buf); } else { int i; printf("Group Name: %-15s #%lu\n" " Members: ", grp->gr_name, (long) grp->gr_gid); if (grp->gr_mem != NULL) { for (i = 0; grp->gr_mem[i]; i++) printf("%s%s", i ? "," : "", grp->gr_mem[i]); } fputs("\n\n", stdout); } return EXIT_SUCCESS; } Index: projects/clang-trunk/usr.sbin/pw/pw_user.c =================================================================== --- projects/clang-trunk/usr.sbin/pw/pw_user.c (revision 283870) +++ projects/clang-trunk/usr.sbin/pw/pw_user.c (revision 283871) @@ -1,1340 +1,1330 @@ /*- * Copyright (C) 1996 * David L. Nugent. All rights reserved. * * Redistribution and use in source and binary forms, with or without * modification, are permitted provided that the following conditions * are met: * 1. Redistributions of source code must retain the above copyright * notice, this list of conditions and the following disclaimer. * 2. Redistributions in binary form must reproduce the above copyright * notice, this list of conditions and the following disclaimer in the * documentation and/or other materials provided with the distribution. * * THIS SOFTWARE IS PROVIDED BY DAVID L. NUGENT AND CONTRIBUTORS ``AS IS'' AND * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE * ARE DISCLAIMED. IN NO EVENT SHALL DAVID L. NUGENT OR CONTRIBUTORS BE LIABLE * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF * SUCH DAMAGE. * */ #ifndef lint static const char rcsid[] = "$FreeBSD$"; #endif /* not lint */ #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include "pw.h" #include "bitmap.h" #define LOGNAMESIZE (MAXLOGNAME-1) static char locked_str[] = "*LOCKED*"; static int print_user(struct passwd * pwd, int pretty, int v7); static uid_t pw_uidpolicy(struct userconf * cnf, struct cargs * args); static uid_t pw_gidpolicy(struct userconf * cnf, struct cargs * args, char *nam, gid_t prefer); static time_t pw_pwdpolicy(struct userconf * cnf, struct cargs * args); static time_t pw_exppolicy(struct userconf * cnf, struct cargs * args); static char *pw_homepolicy(struct userconf * cnf, struct cargs * args, char const * user); static char *pw_shellpolicy(struct userconf * cnf, struct cargs * args, char *newshell); static char *pw_password(struct userconf * cnf, struct cargs * args, char const * user); static char *shell_path(char const * path, char *shells[], char *sh); static void rmat(uid_t uid); static void rmopie(char const * name); /*- * -C config configuration file * -q quiet operation * -n name login name * -u uid user id * -c comment user name/comment * -d directory home directory * -e date account expiry date * -p date password expiry date * -g grp primary group * -G grp1,grp2 additional groups * -m [ -k dir ] create and set up home * -s shell name of login shell * -o duplicate uid ok * -L class user class * -l name new login name * -h fd password filehandle * -H fd encrypted password filehandle * -F force print or add * Setting defaults: * -D set user defaults * -b dir default home root dir * -e period default expiry period * -p period default password change period * -g group default group * -G grp1,grp2.. default additional groups * -L class default login class * -k dir default home skeleton * -s shell default shell * -w method default password method */ int pw_user(struct userconf * cnf, int mode, struct cargs * args) { int rc, edited = 0; char *p = NULL; char *passtmp; struct carg *a_name; struct carg *a_uid; struct carg *arg; struct passwd *pwd = NULL; struct group *grp; struct stat st; char line[_PASSWORD_LEN+1]; FILE *fp; char *dmode_c; void *set = NULL; static struct passwd fakeuser = { NULL, "*", -1, -1, 0, "", "User &", "/nonexistent", "/bin/sh", 0 #if defined(__FreeBSD__) ,0 #endif }; /* * With M_NEXT, we only need to return the * next uid to stdout */ if (mode == M_NEXT) { uid_t next = pw_uidpolicy(cnf, args); if (getarg(args, 'q')) return next; - printf("%ld:", (long)next); + printf("%u:", next); pw_group(cnf, mode, args); return EXIT_SUCCESS; } /* * We can do all of the common legwork here */ if ((arg = getarg(args, 'b')) != NULL) { cnf->home = arg->val; } if ((arg = getarg(args, 'M')) != NULL) { dmode_c = arg->val; if ((set = setmode(dmode_c)) == NULL) errx(EX_DATAERR, "invalid directory creation mode '%s'", dmode_c); cnf->homemode = getmode(set, _DEF_DIRMODE); free(set); } /* * If we'll need to use it or we're updating it, * then create the base home directory if necessary */ if (arg != NULL || getarg(args, 'm') != NULL) { int l = strlen(cnf->home); if (l > 1 && cnf->home[l-1] == '/') /* Shave off any trailing path delimiter */ cnf->home[--l] = '\0'; if (l < 2 || *cnf->home != '/') /* Check for absolute path name */ errx(EX_DATAERR, "invalid base directory for home '%s'", cnf->home); if (stat(cnf->home, &st) == -1) { char dbuf[MAXPATHLEN]; /* * This is a kludge especially for Joerg :) * If the home directory would be created in the root partition, then * we really create it under /usr which is likely to have more space. * But we create a symlink from cnf->home -> "/usr" -> cnf->home */ if (strchr(cnf->home+1, '/') == NULL) { snprintf(dbuf, MAXPATHLEN, "/usr%s", cnf->home); if (mkdir(dbuf, _DEF_DIRMODE) != -1 || errno == EEXIST) { chown(dbuf, 0, 0); /* * Skip first "/" and create symlink: * /home -> usr/home */ symlink(dbuf+1, cnf->home); } /* If this falls, fall back to old method */ } strlcpy(dbuf, cnf->home, sizeof(dbuf)); p = dbuf; if (stat(dbuf, &st) == -1) { while ((p = strchr(p + 1, '/')) != NULL) { *p = '\0'; if (stat(dbuf, &st) == -1) { if (mkdir(dbuf, _DEF_DIRMODE) == -1) goto direrr; chown(dbuf, 0, 0); } else if (!S_ISDIR(st.st_mode)) errx(EX_OSFILE, "'%s' (root home parent) is not a directory", dbuf); *p = '/'; } } if (stat(dbuf, &st) == -1) { if (mkdir(dbuf, _DEF_DIRMODE) == -1) { direrr: err(EX_OSFILE, "mkdir '%s'", dbuf); } chown(dbuf, 0, 0); } } else if (!S_ISDIR(st.st_mode)) errx(EX_OSFILE, "root home `%s' is not a directory", cnf->home); } if ((arg = getarg(args, 'e')) != NULL) cnf->expire_days = atoi(arg->val); if ((arg = getarg(args, 'y')) != NULL) cnf->nispasswd = arg->val; if ((arg = getarg(args, 'p')) != NULL && arg->val) cnf->password_days = atoi(arg->val); if ((arg = getarg(args, 'g')) != NULL) { if (!*(p = arg->val)) /* Handle empty group list specially */ cnf->default_group = ""; else { if ((grp = GETGRNAM(p)) == NULL) { if (!isdigit((unsigned char)*p) || (grp = GETGRGID((gid_t) atoi(p))) == NULL) errx(EX_NOUSER, "group `%s' does not exist", p); } cnf->default_group = newstr(grp->gr_name); } } if ((arg = getarg(args, 'L')) != NULL) cnf->default_class = pw_checkname((u_char *)arg->val, 0); if ((arg = getarg(args, 'G')) != NULL && arg->val) { int i = 0; for (p = strtok(arg->val, ", \t"); p != NULL; p = strtok(NULL, ", \t")) { if ((grp = GETGRNAM(p)) == NULL) { if (!isdigit((unsigned char)*p) || (grp = GETGRGID((gid_t) atoi(p))) == NULL) errx(EX_NOUSER, "group `%s' does not exist", p); } if (extendarray(&cnf->groups, &cnf->numgroups, i + 2) != -1) cnf->groups[i++] = newstr(grp->gr_name); } while (i < cnf->numgroups) cnf->groups[i++] = NULL; } if ((arg = getarg(args, 'k')) != NULL) { if (stat(cnf->dotdir = arg->val, &st) == -1 || !S_ISDIR(st.st_mode)) errx(EX_OSFILE, "skeleton `%s' is not a directory or does not exist", cnf->dotdir); } if ((arg = getarg(args, 's')) != NULL) cnf->shell_default = arg->val; if ((arg = getarg(args, 'w')) != NULL) cnf->default_password = boolean_val(arg->val, cnf->default_password); if (mode == M_ADD && getarg(args, 'D')) { if (getarg(args, 'n') != NULL) errx(EX_DATAERR, "can't combine `-D' with `-n name'"); if ((arg = getarg(args, 'u')) != NULL && (p = strtok(arg->val, ", \t")) != NULL) { if ((cnf->min_uid = (uid_t) atoi(p)) == 0) cnf->min_uid = 1000; if ((p = strtok(NULL, " ,\t")) == NULL || (cnf->max_uid = (uid_t) atoi(p)) < cnf->min_uid) cnf->max_uid = 32000; } if ((arg = getarg(args, 'i')) != NULL && (p = strtok(arg->val, ", \t")) != NULL) { if ((cnf->min_gid = (gid_t) atoi(p)) == 0) cnf->min_gid = 1000; if ((p = strtok(NULL, " ,\t")) == NULL || (cnf->max_gid = (gid_t) atoi(p)) < cnf->min_gid) cnf->max_gid = 32000; } arg = getarg(args, 'C'); if (write_userconfig(arg ? arg->val : NULL)) return EXIT_SUCCESS; - warn("config update"); - return EX_IOERR; + err(EX_IOERR, "config udpate"); } if (mode == M_PRINT && getarg(args, 'a')) { int pretty = getarg(args, 'P') != NULL; int v7 = getarg(args, '7') != NULL; SETPWENT(); while ((pwd = GETPWENT()) != NULL) print_user(pwd, pretty, v7); ENDPWENT(); return EXIT_SUCCESS; } if ((a_name = getarg(args, 'n')) != NULL) pwd = GETPWNAM(pw_checkname((u_char *)a_name->val, 0)); a_uid = getarg(args, 'u'); if (a_uid == NULL) { if (a_name == NULL) errx(EX_DATAERR, "user name or id required"); /* * Determine whether 'n' switch is name or uid - we don't * really don't really care which we have, but we need to * know. */ if (mode != M_ADD && pwd == NULL && strspn(a_name->val, "0123456789") == strlen(a_name->val) && *a_name->val) { (a_uid = a_name)->ch = 'u'; a_name = NULL; } } else { if (strspn(a_uid->val, "0123456789") != strlen(a_uid->val)) errx(EX_USAGE, "-u expects a number"); } /* * Update, delete & print require that the user exists */ if (mode == M_UPDATE || mode == M_DELETE || mode == M_PRINT || mode == M_LOCK || mode == M_UNLOCK) { if (a_name == NULL && pwd == NULL) /* Try harder */ pwd = GETPWUID(atoi(a_uid->val)); if (pwd == NULL) { if (mode == M_PRINT && getarg(args, 'F')) { fakeuser.pw_name = a_name ? a_name->val : "nouser"; - fakeuser.pw_uid = a_uid ? (uid_t) atol(a_uid->val) : -1; + fakeuser.pw_uid = a_uid ? (uid_t) atol(a_uid->val) : (uid_t) -1; return print_user(&fakeuser, getarg(args, 'P') != NULL, getarg(args, '7') != NULL); } if (a_name == NULL) errx(EX_NOUSER, "no such uid `%s'", a_uid->val); errx(EX_NOUSER, "no such user `%s'", a_name->val); } if (a_name == NULL) /* May be needed later */ a_name = addarg(args, 'n', newstr(pwd->pw_name)); /* * The M_LOCK and M_UNLOCK functions simply add or remove * a "*LOCKED*" prefix from in front of the password to * prevent it decoding correctly, and therefore prevents * access. Of course, this only prevents access via * password authentication (not ssh, kerberos or any * other method that does not use the UNIX password) but * that is a known limitation. */ if (mode == M_LOCK) { if (strncmp(pwd->pw_passwd, locked_str, sizeof(locked_str)-1) == 0) errx(EX_DATAERR, "user '%s' is already locked", pwd->pw_name); asprintf(&passtmp, "%s%s", locked_str, pwd->pw_passwd); if (passtmp == NULL) /* disaster */ errx(EX_UNAVAILABLE, "out of memory"); pwd->pw_passwd = passtmp; edited = 1; } else if (mode == M_UNLOCK) { if (strncmp(pwd->pw_passwd, locked_str, sizeof(locked_str)-1) != 0) errx(EX_DATAERR, "user '%s' is not locked", pwd->pw_name); pwd->pw_passwd += sizeof(locked_str)-1; edited = 1; } else if (mode == M_DELETE) { /* * Handle deletions now */ char file[MAXPATHLEN]; char home[MAXPATHLEN]; uid_t uid = pwd->pw_uid; struct group *gr; char grname[LOGNAMESIZE]; if (strcmp(pwd->pw_name, "root") == 0) errx(EX_DATAERR, "cannot remove user 'root'"); if (!PWALTDIR()) { /* * Remove opie record from /etc/opiekeys */ rmopie(pwd->pw_name); /* * Remove crontabs */ snprintf(file, sizeof(file), "/var/cron/tabs/%s", pwd->pw_name); if (access(file, F_OK) == 0) { snprintf(file, sizeof(file), "crontab -u %s -r", pwd->pw_name); system(file); } } /* * Save these for later, since contents of pwd may be * invalidated by deletion */ snprintf(file, sizeof(file), "%s/%s", _PATH_MAILDIR, pwd->pw_name); strlcpy(home, pwd->pw_dir, sizeof(home)); gr = GETGRGID(pwd->pw_gid); if (gr != NULL) strlcpy(grname, gr->gr_name, LOGNAMESIZE); else grname[0] = '\0'; rc = delpwent(pwd); if (rc == -1) err(EX_IOERR, "user '%s' does not exist", pwd->pw_name); - else if (rc != 0) { - warn("passwd update"); - return EX_IOERR; - } + else if (rc != 0) + err(EX_IOERR, "passwd update"); if (cnf->nispasswd && *cnf->nispasswd=='/') { rc = delnispwent(cnf->nispasswd, a_name->val); if (rc == -1) warnx("WARNING: user '%s' does not exist in NIS passwd", pwd->pw_name); else if (rc != 0) warn("WARNING: NIS passwd update"); /* non-fatal */ } grp = GETGRNAM(a_name->val); if (grp != NULL && (grp->gr_mem == NULL || *grp->gr_mem == NULL) && strcmp(a_name->val, grname) == 0) delgrent(GETGRNAM(a_name->val)); SETGRENT(); while ((grp = GETGRENT()) != NULL) { int i, j; char group[MAXLOGNAME]; if (grp->gr_mem != NULL) { for (i = 0; grp->gr_mem[i] != NULL; i++) { if (!strcmp(grp->gr_mem[i], a_name->val)) { for (j = i; grp->gr_mem[j] != NULL; j++) grp->gr_mem[j] = grp->gr_mem[j+1]; strlcpy(group, grp->gr_name, MAXLOGNAME); chggrent(group, grp); } } } } ENDGRENT(); - pw_log(cnf, mode, W_USER, "%s(%ld) account removed", a_name->val, (long) uid); + pw_log(cnf, mode, W_USER, "%s(%u) account removed", a_name->val, uid); if (!PWALTDIR()) { /* * Remove mail file */ remove(file); /* * Remove at jobs */ if (getpwuid(uid) == NULL) rmat(uid); /* * Remove home directory and contents */ if (getarg(args, 'r') != NULL && *home == '/' && getpwuid(uid) == NULL) { if (stat(home, &st) != -1) { rm_r(home, uid); - pw_log(cnf, mode, W_USER, "%s(%ld) home '%s' %sremoved", - a_name->val, (long) uid, home, + pw_log(cnf, mode, W_USER, "%s(%u) home '%s' %sremoved", + a_name->val, uid, home, stat(home, &st) == -1 ? "" : "not completely "); } } } return EXIT_SUCCESS; } else if (mode == M_PRINT) return print_user(pwd, getarg(args, 'P') != NULL, getarg(args, '7') != NULL); /* * The rest is edit code */ if ((arg = getarg(args, 'l')) != NULL) { if (strcmp(pwd->pw_name, "root") == 0) errx(EX_DATAERR, "can't rename `root' account"); pwd->pw_name = pw_checkname((u_char *)arg->val, 0); edited = 1; } if ((arg = getarg(args, 'u')) != NULL && isdigit((unsigned char)*arg->val)) { pwd->pw_uid = (uid_t) atol(arg->val); edited = 1; if (pwd->pw_uid != 0 && strcmp(pwd->pw_name, "root") == 0) errx(EX_DATAERR, "can't change uid of `root' account"); if (pwd->pw_uid == 0 && strcmp(pwd->pw_name, "root") != 0) warnx("WARNING: account `%s' will have a uid of 0 (superuser access!)", pwd->pw_name); } if ((arg = getarg(args, 'g')) != NULL && pwd->pw_uid != 0) { /* Already checked this */ gid_t newgid = (gid_t) GETGRNAM(cnf->default_group)->gr_gid; if (newgid != pwd->pw_gid) { edited = 1; pwd->pw_gid = newgid; } } if ((arg = getarg(args, 'p')) != NULL) { if (*arg->val == '\0' || strcmp(arg->val, "0") == 0) { if (pwd->pw_change != 0) { pwd->pw_change = 0; edited = 1; } } else { time_t now = time(NULL); time_t expire = parse_date(now, arg->val); if (pwd->pw_change != expire) { pwd->pw_change = expire; edited = 1; } } } if ((arg = getarg(args, 'e')) != NULL) { if (*arg->val == '\0' || strcmp(arg->val, "0") == 0) { if (pwd->pw_expire != 0) { pwd->pw_expire = 0; edited = 1; } } else { time_t now = time(NULL); time_t expire = parse_date(now, arg->val); if (pwd->pw_expire != expire) { pwd->pw_expire = expire; edited = 1; } } } if ((arg = getarg(args, 's')) != NULL) { char *shell = shell_path(cnf->shelldir, cnf->shells, arg->val); if (shell == NULL) shell = ""; if (strcmp(shell, pwd->pw_shell) != 0) { pwd->pw_shell = shell; edited = 1; } } if (getarg(args, 'L')) { if (cnf->default_class == NULL) cnf->default_class = ""; if (strcmp(pwd->pw_class, cnf->default_class) != 0) { pwd->pw_class = cnf->default_class; edited = 1; } } if ((arg = getarg(args, 'd')) != NULL) { if (strcmp(pwd->pw_dir, arg->val)) edited = 1; if (stat(pwd->pw_dir = arg->val, &st) == -1) { if (getarg(args, 'm') == NULL && strcmp(pwd->pw_dir, "/nonexistent") != 0) warnx("WARNING: home `%s' does not exist", pwd->pw_dir); } else if (!S_ISDIR(st.st_mode)) warnx("WARNING: home `%s' is not a directory", pwd->pw_dir); } if ((arg = getarg(args, 'w')) != NULL && getarg(args, 'h') == NULL && getarg(args, 'H') == NULL) { login_cap_t *lc; lc = login_getpwclass(pwd); if (lc == NULL || login_setcryptfmt(lc, "sha512", NULL) == NULL) warn("setting crypt(3) format"); login_close(lc); pwd->pw_passwd = pw_password(cnf, args, pwd->pw_name); edited = 1; } } else { login_cap_t *lc; /* * Add code */ if (a_name == NULL) /* Required */ errx(EX_DATAERR, "login name required"); else if ((pwd = GETPWNAM(a_name->val)) != NULL) /* Exists */ errx(EX_DATAERR, "login name `%s' already exists", a_name->val); /* * Now, set up defaults for a new user */ pwd = &fakeuser; pwd->pw_name = a_name->val; pwd->pw_class = cnf->default_class ? cnf->default_class : ""; pwd->pw_uid = pw_uidpolicy(cnf, args); pwd->pw_gid = pw_gidpolicy(cnf, args, pwd->pw_name, (gid_t) pwd->pw_uid); pwd->pw_change = pw_pwdpolicy(cnf, args); pwd->pw_expire = pw_exppolicy(cnf, args); pwd->pw_dir = pw_homepolicy(cnf, args, pwd->pw_name); pwd->pw_shell = pw_shellpolicy(cnf, args, NULL); lc = login_getpwclass(pwd); if (lc == NULL || login_setcryptfmt(lc, "sha512", NULL) == NULL) warn("setting crypt(3) format"); login_close(lc); pwd->pw_passwd = pw_password(cnf, args, pwd->pw_name); edited = 1; if (pwd->pw_uid == 0 && strcmp(pwd->pw_name, "root") != 0) warnx("WARNING: new account `%s' has a uid of 0 (superuser access!)", pwd->pw_name); } /* * Shared add/edit code */ if ((arg = getarg(args, 'c')) != NULL) { char *gecos = pw_checkname((u_char *)arg->val, 1); if (strcmp(pwd->pw_gecos, gecos) != 0) { pwd->pw_gecos = gecos; edited = 1; } } if ((arg = getarg(args, 'h')) != NULL || (arg = getarg(args, 'H')) != NULL) { if (strcmp(arg->val, "-") == 0) { if (!pwd->pw_passwd || *pwd->pw_passwd != '*') { pwd->pw_passwd = "*"; /* No access */ edited = 1; } } else { int fd = atoi(arg->val); int precrypt = (arg->ch == 'H'); int b; int istty = isatty(fd); struct termios t; login_cap_t *lc; if (istty) { if (tcgetattr(fd, &t) == -1) istty = 0; else { struct termios n = t; /* Disable echo */ n.c_lflag &= ~(ECHO); tcsetattr(fd, TCSANOW, &n); printf("%s%spassword for user %s:", (mode == M_UPDATE) ? "new " : "", precrypt ? "encrypted " : "", pwd->pw_name); fflush(stdout); } } b = read(fd, line, sizeof(line) - 1); if (istty) { /* Restore state */ tcsetattr(fd, TCSANOW, &t); fputc('\n', stdout); fflush(stdout); } - if (b < 0) { - warn("-%c file descriptor", precrypt ? 'H' : - 'h'); - return EX_IOERR; - } + if (b < 0) + err(EX_IOERR, "-%c file descriptor", + precrypt ? 'H' : 'h'); line[b] = '\0'; if ((p = strpbrk(line, "\r\n")) != NULL) *p = '\0'; if (!*line) errx(EX_DATAERR, "empty password read on file descriptor %d", fd); if (precrypt) { if (strchr(line, ':') != NULL) return EX_DATAERR; pwd->pw_passwd = line; } else { lc = login_getpwclass(pwd); if (lc == NULL || login_setcryptfmt(lc, "sha512", NULL) == NULL) warn("setting crypt(3) format"); login_close(lc); pwd->pw_passwd = pw_pwcrypt(line); } edited = 1; } } /* * Special case: -N only displays & exits */ if (getarg(args, 'N') != NULL) return print_user(pwd, getarg(args, 'P') != NULL, getarg(args, '7') != NULL); if (mode == M_ADD) { edited = 1; /* Always */ rc = addpwent(pwd); - if (rc == -1) { - warnx("user '%s' already exists", pwd->pw_name); - return EX_IOERR; - } else if (rc != 0) { - warn("passwd file update"); - return EX_IOERR; - } + if (rc == -1) + errx(EX_IOERR, "user '%s' already exists", + pwd->pw_name); + else if (rc != 0) + err(EX_IOERR, "passwd file update"); if (cnf->nispasswd && *cnf->nispasswd=='/') { rc = addnispwent(cnf->nispasswd, pwd); if (rc == -1) warnx("User '%s' already exists in NIS passwd", pwd->pw_name); else warn("NIS passwd update"); /* NOTE: we treat NIS-only update errors as non-fatal */ } } else if (mode == M_UPDATE || mode == M_LOCK || mode == M_UNLOCK) { if (edited) { /* Only updated this if required */ rc = chgpwent(a_name->val, pwd); - if (rc == -1) { - warnx("user '%s' does not exist (NIS?)", pwd->pw_name); - return EX_IOERR; - } else if (rc != 0) { - warn("passwd file update"); - return EX_IOERR; - } + if (rc == -1) + errx(EX_IOERR, "user '%s' does not exist (NIS?)", pwd->pw_name); + else if (rc != 0) + err(EX_IOERR, "passwd file update"); if ( cnf->nispasswd && *cnf->nispasswd=='/') { rc = chgnispwent(cnf->nispasswd, a_name->val, pwd); if (rc == -1) warn("User '%s' not found in NIS passwd", pwd->pw_name); else warn("NIS passwd update"); /* NOTE: NIS-only update errors are not fatal */ } } } /* * Ok, user is created or changed - now edit group file */ if (mode == M_ADD || getarg(args, 'G') != NULL) { int i, j; /* First remove the user from all group */ SETGRENT(); while ((grp = GETGRENT()) != NULL) { char group[MAXLOGNAME]; if (grp->gr_mem == NULL) continue; for (i = 0; grp->gr_mem[i] != NULL; i++) { if (strcmp(grp->gr_mem[i] , pwd->pw_name) != 0) continue; for (j = i; grp->gr_mem[j] != NULL ; j++) grp->gr_mem[j] = grp->gr_mem[j+1]; strlcpy(group, grp->gr_name, MAXLOGNAME); chggrent(group, grp); } } ENDGRENT(); /* now add to group where needed */ for (i = 0; cnf->groups[i] != NULL; i++) { grp = GETGRNAM(cnf->groups[i]); grp = gr_add(grp, pwd->pw_name); /* * grp can only be NULL in 2 cases: * - the new member is already a member * - a problem with memory occurs * in both cases we want to skip now. */ if (grp == NULL) continue; chggrent(cnf->groups[i], grp); free(grp); } } /* go get a current version of pwd */ pwd = GETPWNAM(a_name->val); if (pwd == NULL) { /* This will fail when we rename, so special case that */ if (mode == M_UPDATE && (arg = getarg(args, 'l')) != NULL) { a_name->val = arg->val; /* update new name */ pwd = GETPWNAM(a_name->val); /* refetch renamed rec */ } } if (pwd == NULL) /* can't go on without this */ errx(EX_NOUSER, "user '%s' disappeared during update", a_name->val); grp = GETGRGID(pwd->pw_gid); - pw_log(cnf, mode, W_USER, "%s(%ld):%s(%ld):%s:%s:%s", - pwd->pw_name, (long) pwd->pw_uid, - grp ? grp->gr_name : "unknown", (long) (grp ? grp->gr_gid : -1), + pw_log(cnf, mode, W_USER, "%s(%u):%s(%u):%s:%s:%s", + pwd->pw_name, pwd->pw_uid, + grp ? grp->gr_name : "unknown", (grp ? grp->gr_gid : (uid_t)-1), pwd->pw_gecos, pwd->pw_dir, pwd->pw_shell); /* * If adding, let's touch and chown the user's mail file. This is not * strictly necessary under BSD with a 0755 maildir but it also * doesn't hurt anything to create the empty mailfile */ if (mode == M_ADD) { if (!PWALTDIR()) { snprintf(line, sizeof(line), "%s/%s", _PATH_MAILDIR, pwd->pw_name); close(open(line, O_RDWR | O_CREAT, 0600)); /* Preserve contents & * mtime */ chown(line, pwd->pw_uid, pwd->pw_gid); } } /* * Let's create and populate the user's home directory. Note * that this also `works' for editing users if -m is used, but * existing files will *not* be overwritten. */ if (!PWALTDIR() && getarg(args, 'm') != NULL && pwd->pw_dir && *pwd->pw_dir == '/' && pwd->pw_dir[1]) { copymkdir(pwd->pw_dir, cnf->dotdir, cnf->homemode, pwd->pw_uid, pwd->pw_gid); - pw_log(cnf, mode, W_USER, "%s(%ld) home %s made", - pwd->pw_name, (long) pwd->pw_uid, pwd->pw_dir); + pw_log(cnf, mode, W_USER, "%s(%u) home %s made", + pwd->pw_name, pwd->pw_uid, pwd->pw_dir); } /* * Finally, send mail to the new user as well, if we are asked to */ if (mode == M_ADD && !PWALTDIR() && cnf->newmail && *cnf->newmail && (fp = fopen(cnf->newmail, "r")) != NULL) { FILE *pfp = popen(_PATH_SENDMAIL " -t", "w"); if (pfp == NULL) warn("sendmail"); else { fprintf(pfp, "From: root\n" "To: %s\n" "Subject: Welcome!\n\n", pwd->pw_name); while (fgets(line, sizeof(line), fp) != NULL) { /* Do substitutions? */ fputs(line, pfp); } pclose(pfp); - pw_log(cnf, mode, W_USER, "%s(%ld) new user mail sent", - pwd->pw_name, (long) pwd->pw_uid); + pw_log(cnf, mode, W_USER, "%s(%u) new user mail sent", + pwd->pw_name, pwd->pw_uid); } fclose(fp); } return EXIT_SUCCESS; } static uid_t pw_uidpolicy(struct userconf * cnf, struct cargs * args) { struct passwd *pwd; uid_t uid = (uid_t) - 1; struct carg *a_uid = getarg(args, 'u'); /* * Check the given uid, if any */ if (a_uid != NULL) { uid = (uid_t) atol(a_uid->val); if ((pwd = GETPWUID(uid)) != NULL && getarg(args, 'o') == NULL) - errx(EX_DATAERR, "uid `%ld' has already been allocated", (long) pwd->pw_uid); + errx(EX_DATAERR, "uid `%u' has already been allocated", pwd->pw_uid); } else { struct bitmap bm; /* * We need to allocate the next available uid under one of * two policies a) Grab the first unused uid b) Grab the * highest possible unused uid */ if (cnf->min_uid >= cnf->max_uid) { /* Sanity * claus^H^H^H^Hheck */ cnf->min_uid = 1000; cnf->max_uid = 32000; } bm = bm_alloc(cnf->max_uid - cnf->min_uid + 1); /* * Now, let's fill the bitmap from the password file */ SETPWENT(); while ((pwd = GETPWENT()) != NULL) if (pwd->pw_uid >= (uid_t) cnf->min_uid && pwd->pw_uid <= (uid_t) cnf->max_uid) bm_setbit(&bm, pwd->pw_uid - cnf->min_uid); ENDPWENT(); /* * Then apply the policy, with fallback to reuse if necessary */ if (cnf->reuse_uids || (uid = (uid_t) (bm_lastset(&bm) + cnf->min_uid + 1)) > cnf->max_uid) uid = (uid_t) (bm_firstunset(&bm) + cnf->min_uid); /* * Another sanity check */ if (uid < cnf->min_uid || uid > cnf->max_uid) errx(EX_SOFTWARE, "unable to allocate a new uid - range fully used"); bm_dealloc(&bm); } return uid; } static uid_t pw_gidpolicy(struct userconf * cnf, struct cargs * args, char *nam, gid_t prefer) { struct group *grp; gid_t gid = (uid_t) - 1; struct carg *a_gid = getarg(args, 'g'); /* * If no arg given, see if default can help out */ if (a_gid == NULL && cnf->default_group && *cnf->default_group) a_gid = addarg(args, 'g', cnf->default_group); /* * Check the given gid, if any */ SETGRENT(); if (a_gid != NULL) { if ((grp = GETGRNAM(a_gid->val)) == NULL) { gid = (gid_t) atol(a_gid->val); if ((gid == 0 && !isdigit((unsigned char)*a_gid->val)) || (grp = GETGRGID(gid)) == NULL) errx(EX_NOUSER, "group `%s' is not defined", a_gid->val); } gid = grp->gr_gid; } else if ((grp = GETGRNAM(nam)) != NULL && (grp->gr_mem == NULL || grp->gr_mem[0] == NULL)) { gid = grp->gr_gid; /* Already created? Use it anyway... */ } else { struct cargs grpargs; char tmp[32]; LIST_INIT(&grpargs); addarg(&grpargs, 'n', nam); /* * We need to auto-create a group with the user's name. We * can send all the appropriate output to our sister routine * bit first see if we can create a group with gid==uid so we * can keep the user and group ids in sync. We purposely do * NOT check the gid range if we can force the sync. If the * user's name dups an existing group, then the group add * function will happily handle that case for us and exit. */ if (GETGRGID(prefer) == NULL) { snprintf(tmp, sizeof(tmp), "%u", prefer); addarg(&grpargs, 'g', tmp); } if (getarg(args, 'N')) { addarg(&grpargs, 'N', NULL); addarg(&grpargs, 'q', NULL); gid = pw_group(cnf, M_NEXT, &grpargs); } else { pw_group(cnf, M_ADD, &grpargs); if ((grp = GETGRNAM(nam)) != NULL) gid = grp->gr_gid; } a_gid = LIST_FIRST(&grpargs); while (a_gid != NULL) { struct carg *t = LIST_NEXT(a_gid, list); LIST_REMOVE(a_gid, list); a_gid = t; } } ENDGRENT(); return gid; } static time_t pw_pwdpolicy(struct userconf * cnf, struct cargs * args) { time_t result = 0; time_t now = time(NULL); struct carg *arg = getarg(args, 'p'); if (arg != NULL) { if ((result = parse_date(now, arg->val)) == now) errx(EX_DATAERR, "invalid date/time `%s'", arg->val); } else if (cnf->password_days > 0) result = now + ((long) cnf->password_days * 86400L); return result; } static time_t pw_exppolicy(struct userconf * cnf, struct cargs * args) { time_t result = 0; time_t now = time(NULL); struct carg *arg = getarg(args, 'e'); if (arg != NULL) { if ((result = parse_date(now, arg->val)) == now) errx(EX_DATAERR, "invalid date/time `%s'", arg->val); } else if (cnf->expire_days > 0) result = now + ((long) cnf->expire_days * 86400L); return result; } static char * pw_homepolicy(struct userconf * cnf, struct cargs * args, char const * user) { struct carg *arg = getarg(args, 'd'); static char home[128]; if (arg) return (arg->val); if (cnf->home == NULL || *cnf->home == '\0') errx(EX_CONFIG, "no base home directory set"); snprintf(home, sizeof(home), "%s/%s", cnf->home, user); return (home); } static char * shell_path(char const * path, char *shells[], char *sh) { if (sh != NULL && (*sh == '/' || *sh == '\0')) return sh; /* specified full path or forced none */ else { char *p; char paths[_UC_MAXLINE]; /* * We need to search paths */ strlcpy(paths, path, sizeof(paths)); for (p = strtok(paths, ": \t\r\n"); p != NULL; p = strtok(NULL, ": \t\r\n")) { int i; static char shellpath[256]; if (sh != NULL) { snprintf(shellpath, sizeof(shellpath), "%s/%s", p, sh); if (access(shellpath, X_OK) == 0) return shellpath; } else for (i = 0; i < _UC_MAXSHELLS && shells[i] != NULL; i++) { snprintf(shellpath, sizeof(shellpath), "%s/%s", p, shells[i]); if (access(shellpath, X_OK) == 0) return shellpath; } } if (sh == NULL) errx(EX_OSFILE, "can't find shell `%s' in shell paths", sh); errx(EX_CONFIG, "no default shell available or defined"); return NULL; } } static char * pw_shellpolicy(struct userconf * cnf, struct cargs * args, char *newshell) { char *sh = newshell; struct carg *arg = getarg(args, 's'); if (newshell == NULL && arg != NULL) sh = arg->val; return shell_path(cnf->shelldir, cnf->shells, sh ? sh : cnf->shell_default); } #define SALTSIZE 32 static char const chars[] = "0123456789abcdefghijklmnopqrstuvwxyzABCDEFGHIJKLMNOPQRSTUVWXYZ./"; char * pw_pwcrypt(char *password) { int i; char salt[SALTSIZE + 1]; char *cryptpw; static char buf[256]; /* * Calculate a salt value */ for (i = 0; i < SALTSIZE; i++) salt[i] = chars[arc4random_uniform(sizeof(chars) - 1)]; salt[SALTSIZE] = '\0'; cryptpw = crypt(password, salt); if (cryptpw == NULL) errx(EX_CONFIG, "crypt(3) failure"); return strcpy(buf, cryptpw); } static char * pw_password(struct userconf * cnf, struct cargs * args, char const * user) { int i, l; char pwbuf[32]; switch (cnf->default_password) { case -1: /* Random password */ l = (arc4random() % 8 + 8); /* 8 - 16 chars */ for (i = 0; i < l; i++) pwbuf[i] = chars[arc4random_uniform(sizeof(chars)-1)]; pwbuf[i] = '\0'; /* * We give this information back to the user */ if (getarg(args, 'h') == NULL && getarg(args, 'H') == NULL && getarg(args, 'N') == NULL) { if (isatty(STDOUT_FILENO)) printf("Password for '%s' is: ", user); printf("%s\n", pwbuf); fflush(stdout); } break; case -2: /* No password at all! */ return ""; case 0: /* No login - default */ default: return "*"; case 1: /* user's name */ strlcpy(pwbuf, user, sizeof(pwbuf)); break; } return pw_pwcrypt(pwbuf); } static int print_user(struct passwd * pwd, int pretty, int v7) { if (!pretty) { char *buf; if (!v7) pwd->pw_passwd = (pwd->pw_passwd == NULL) ? "" : "*"; buf = v7 ? pw_make_v7(pwd) : pw_make(pwd); printf("%s\n", buf); free(buf); } else { int j; char *p; struct group *grp = GETGRGID(pwd->pw_gid); char uname[60] = "User &", office[60] = "[None]", wphone[60] = "[None]", hphone[60] = "[None]"; char acexpire[32] = "[None]", pwexpire[32] = "[None]"; struct tm * tptr; if ((p = strtok(pwd->pw_gecos, ",")) != NULL) { strlcpy(uname, p, sizeof(uname)); if ((p = strtok(NULL, ",")) != NULL) { strlcpy(office, p, sizeof(office)); if ((p = strtok(NULL, ",")) != NULL) { strlcpy(wphone, p, sizeof(wphone)); if ((p = strtok(NULL, "")) != NULL) { strlcpy(hphone, p, sizeof(hphone)); } } } } /* * Handle '&' in gecos field */ if ((p = strchr(uname, '&')) != NULL) { int l = strlen(pwd->pw_name); int m = strlen(p); memmove(p + l, p + 1, m); memmove(p, pwd->pw_name, l); *p = (char) toupper((unsigned char)*p); } if (pwd->pw_expire > (time_t)0 && (tptr = localtime(&pwd->pw_expire)) != NULL) strftime(acexpire, sizeof acexpire, "%c", tptr); if (pwd->pw_change > (time_t)0 && (tptr = localtime(&pwd->pw_change)) != NULL) strftime(pwexpire, sizeof pwexpire, "%c", tptr); - printf("Login Name: %-15s #%-12ld Group: %-15s #%ld\n" + printf("Login Name: %-15s #%-12u Group: %-15s #%u\n" " Full Name: %s\n" " Home: %-26.26s Class: %s\n" " Shell: %-26.26s Office: %s\n" "Work Phone: %-26.26s Home Phone: %s\n" "Acc Expire: %-26.26s Pwd Expire: %s\n", - pwd->pw_name, (long) pwd->pw_uid, - grp ? grp->gr_name : "(invalid)", (long) pwd->pw_gid, + pwd->pw_name, pwd->pw_uid, + grp ? grp->gr_name : "(invalid)", pwd->pw_gid, uname, pwd->pw_dir, pwd->pw_class, pwd->pw_shell, office, wphone, hphone, acexpire, pwexpire); SETGRENT(); j = 0; while ((grp=GETGRENT()) != NULL) { int i = 0; if (grp->gr_mem != NULL) { while (grp->gr_mem[i] != NULL) { if (strcmp(grp->gr_mem[i], pwd->pw_name)==0) { printf(j++ == 0 ? " Groups: %s" : ",%s", grp->gr_name); break; } ++i; } } } ENDGRENT(); printf("%s", j ? "\n" : ""); } return EXIT_SUCCESS; } char * pw_checkname(u_char *name, int gecos) { char showch[8]; u_char const *badchars, *ch, *showtype; int reject; ch = name; reject = 0; if (gecos) { /* See if the name is valid as a gecos (comment) field. */ badchars = ":!@"; showtype = "gecos field"; } else { /* See if the name is valid as a userid or group. */ badchars = " ,\t:+&#%$^()!@~*?<>=|\\/\""; showtype = "userid/group name"; /* Userids and groups can not have a leading '-'. */ if (*ch == '-') reject = 1; } if (!reject) { while (*ch) { if (strchr(badchars, *ch) != NULL || *ch < ' ' || *ch == 127) { reject = 1; break; } /* 8-bit characters are only allowed in GECOS fields */ if (!gecos && (*ch & 0x80)) { reject = 1; break; } ch++; } } /* * A `$' is allowed as the final character for userids and groups, * mainly for the benefit of samba. */ if (reject && !gecos) { if (*ch == '$' && *(ch + 1) == '\0') { reject = 0; ch++; } } if (reject) { snprintf(showch, sizeof(showch), (*ch >= ' ' && *ch < 127) ? "`%c'" : "0x%02x", *ch); errx(EX_DATAERR, "invalid character %s at position %td in %s", showch, (ch - name), showtype); } if (!gecos && (ch - name) > LOGNAMESIZE) errx(EX_DATAERR, "name too long `%s' (max is %d)", name, LOGNAMESIZE); return (char *)name; } static void rmat(uid_t uid) { DIR *d = opendir("/var/at/jobs"); if (d != NULL) { struct dirent *e; while ((e = readdir(d)) != NULL) { struct stat st; if (strncmp(e->d_name, ".lock", 5) != 0 && stat(e->d_name, &st) == 0 && !S_ISDIR(st.st_mode) && st.st_uid == uid) { char tmp[MAXPATHLEN]; snprintf(tmp, sizeof(tmp), "/usr/bin/atrm %s", e->d_name); system(tmp); } } closedir(d); } } static void rmopie(char const * name) { static const char etcopie[] = "/etc/opiekeys"; FILE *fp = fopen(etcopie, "r+"); if (fp != NULL) { char tmp[1024]; off_t atofs = 0; int length = strlen(name); while (fgets(tmp, sizeof tmp, fp) != NULL) { if (strncmp(name, tmp, length) == 0 && tmp[length]==' ') { if (fseek(fp, atofs, SEEK_SET) == 0) { fwrite("#", 1, 1, fp); /* Comment username out */ } break; } atofs = ftell(fp); } /* * If we got an error of any sort, don't update! */ fclose(fp); } } Index: projects/clang-trunk/usr.sbin/pw/pwupd.c =================================================================== --- projects/clang-trunk/usr.sbin/pw/pwupd.c (revision 283870) +++ projects/clang-trunk/usr.sbin/pw/pwupd.c (revision 283871) @@ -1,170 +1,167 @@ /*- * Copyright (C) 1996 * David L. Nugent. All rights reserved. * * Redistribution and use in source and binary forms, with or without * modification, are permitted provided that the following conditions * are met: * 1. Redistributions of source code must retain the above copyright * notice, this list of conditions and the following disclaimer. * 2. Redistributions in binary form must reproduce the above copyright * notice, this list of conditions and the following disclaimer in the * documentation and/or other materials provided with the distribution. * * THIS SOFTWARE IS PROVIDED BY DAVID L. NUGENT AND CONTRIBUTORS ``AS IS'' AND * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE * ARE DISCLAIMED. IN NO EVENT SHALL DAVID L. NUGENT OR CONTRIBUTORS BE LIABLE * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF * SUCH DAMAGE. */ #ifndef lint static const char rcsid[] = "$FreeBSD$"; #endif /* not lint */ #include #include #include #include #include #include #include #include #include #include #include #include #include "pwupd.h" static char pathpwd[] = _PATH_PWD; static char * pwpath = pathpwd; int setpwdir(const char * dir) { if (dir == NULL) return (-1); else pwpath = strdup(dir); if (pwpath == NULL) return (-1); return (0); } char * getpwpath(char const * file) { static char pathbuf[MAXPATHLEN]; snprintf(pathbuf, sizeof pathbuf, "%s/%s", pwpath, file); return (pathbuf); } static int pwdb_check(void) { int i = 0; pid_t pid; char *args[10]; args[i++] = _PATH_PWD_MKDB; args[i++] = "-C"; if (pwpath != pathpwd) { args[i++] = "-d"; args[i++] = pwpath; } args[i++] = getpwpath(_MASTERPASSWD); args[i] = NULL; if ((pid = fork()) == -1) /* Error (errno set) */ i = errno; else if (pid == 0) { /* Child */ execv(args[0], args); _exit(1); } else { /* Parent */ waitpid(pid, &i, 0); if (WEXITSTATUS(i)) i = EIO; } return (i); } static int pw_update(struct passwd * pwd, char const * user) { struct passwd *pw = NULL; struct passwd *old_pw = NULL; int rc, pfd, tfd; if ((rc = pwdb_check()) != 0) return (rc); if (pwd != NULL) pw = pw_dup(pwd); if (user != NULL) old_pw = GETPWNAM(user); if (pw_init(pwpath, NULL)) err(1, "pw_init()"); if ((pfd = pw_lock()) == -1) { pw_fini(); err(1, "pw_lock()"); } if ((tfd = pw_tmp(-1)) == -1) { pw_fini(); err(1, "pw_tmp()"); } if (pw_copy(pfd, tfd, pw, old_pw) == -1) { pw_fini(); err(1, "pw_copy()"); } /* * in case of deletion of a user, the whole database * needs to be regenerated */ if (pw_mkdb(pw != NULL ? pw->pw_name : NULL) == -1) { pw_fini(); err(1, "pw_mkdb()"); } free(pw); pw_fini(); return (0); } int addpwent(struct passwd * pwd) { return (pw_update(pwd, NULL)); } int chgpwent(char const * login, struct passwd * pwd) { return (pw_update(pwd, login)); } int delpwent(struct passwd * pwd) { - char login[MAXLOGNAME]; - - strlcpy(login, pwd->pw_name, MAXLOGNAME); - return (pw_update(NULL, login)); + return (pw_update(NULL, pwd->pw_name)); } Index: projects/clang-trunk/usr.sbin/yppoll/yppoll.8 =================================================================== --- projects/clang-trunk/usr.sbin/yppoll/yppoll.8 (revision 283870) +++ projects/clang-trunk/usr.sbin/yppoll/yppoll.8 (revision 283871) @@ -1,77 +1,77 @@ .\" Copyright (c) 1996 The NetBSD Foundation, Inc. .\" All rights reserved. .\" .\" This code is derived from software contributed to The NetBSD Foundation .\" by Jason R. Thorpe. .\" .\" Redistribution and use in source and binary forms, with or without .\" modification, are permitted provided that the following conditions .\" are met: .\" 1. Redistributions of source code must retain the above copyright .\" notice, this list of conditions and the following disclaimer. .\" 2. Redistributions in binary form must reproduce the above copyright .\" notice, this list of conditions and the following disclaimer in the .\" documentation and/or other materials provided with the distribution. .\" .\" THIS SOFTWARE IS PROVIDED BY THE NETBSD FOUNDATION, INC. AND CONTRIBUTORS .\" ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED .\" TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR .\" PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE FOUNDATION OR CONTRIBUTORS .\" BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR .\" CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF .\" SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS .\" INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN .\" CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) .\" ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE .\" POSSIBILITY OF SUCH DAMAGE. .\" .\" $FreeBSD$ .\" -.Dd October 25, 1994 +.Dd May 31, 2015 .Dt YPPOLL 8 .Os .Sh NAME .Nm yppoll .Nd ask version of YP map from YP server .Sh SYNOPSIS .Nm -.Op Fl h Ar host +.\".Op Fl h Ar host .Op Fl d Ar domain .Ar mapname .Sh DESCRIPTION The .Nm utility asks a YP server process for the order number and which host is the master server for .Ar mapname . .Pp The options are as follows: .Bl -tag -width indent -.It Fl h Ar host -Ask the YP server process running on -.Ar host -for information about -.Ar mapname . -If -.Ar host -is not specified, the server polled is the default server returned by -.Xr ypwhich 1 . +.\".It Fl h Ar host +.\"Ask the YP server process running on +.\".Ar host +.\"for information about +.\".Ar mapname . +.\"If +.\".Ar host +.\"is not specified, the server polled is the default server returned by +.\".Xr ypwhich 1 . .It Fl d Ar domain Use the YP domain .Ar domain instead of the default domain as returned by .Xr domainname 1 . .El .Sh SEE ALSO .Xr domainname 1 , .Xr ypcat 1 , .Xr ypmatch 1 , .Xr ypwhich 1 , .Xr yp 8 , .Xr ypbind 8 , .Xr ypset 8 .Sh AUTHORS .An Theo De Raadt and .An John Brezak Index: projects/clang-trunk/usr.sbin/yppoll/yppoll.c =================================================================== --- projects/clang-trunk/usr.sbin/yppoll/yppoll.c (revision 283870) +++ projects/clang-trunk/usr.sbin/yppoll/yppoll.c (revision 283871) @@ -1,97 +1,107 @@ /* * Copyright (c) 1992/3 Theo de Raadt * Copyright (c) 1992/3 John Brezak * All rights reserved. * * Redistribution and use in source and binary forms, with or without * modification, are permitted provided that the following conditions * are met: * 1. Redistributions of source code must retain the above copyright * notice, this list of conditions and the following disclaimer. * 2. Redistributions in binary form must reproduce the above copyright * notice, this list of conditions and the following disclaimer in the * documentation and/or other materials provided with the distribution. * 3. The name of the author may not be used to endorse or promote * products derived from this software without specific prior written * permission. * * THIS SOFTWARE IS PROVIDED BY THE AUTHOR ``AS IS'' AND ANY EXPRESS * OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED * WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE * ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR ANY * DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF * SUCH DAMAGE. */ #include __FBSDID("$FreeBSD$"); #include #include #include #include #include #include #include #include #include #include #include #include #include static void usage(void) { +#if 0 fprintf(stderr, "usage: yppoll [-h host] [-d domainname] mapname\n"); +#else + fprintf(stderr, "usage: yppoll [-d domainname] mapname\n"); +#endif exit(1); } int main(int argc, char *argv[]) { char *domainname; +#if 0 char *hostname = "localhost"; +#endif char *inmap, *master; int order; int c, r; time_t t; yp_get_default_domain(&domainname); while ((c = getopt(argc, argv, "h:d:")) != -1) switch (c) { case 'd': domainname = optarg; break; case 'h': +#if 0 hostname = optarg; +#else + /* does nothing */ +#endif break; case '?': usage(); /*NOTREACHED*/ } if (optind + 1 != argc) usage(); inmap = argv[optind]; r = yp_order(domainname, inmap, &order); if (r != 0) errx(1, "no such map %s. Reason: %s", inmap, yperr_string(r)); t = _int_to_time(order); printf("Map %s has order number %d. %s", inmap, order, ctime(&t)); r = yp_master(domainname, inmap, &master); if (r != 0) errx(1, "no such map %s. Reason: %s", inmap, yperr_string(r)); printf("The master server is %s.\n", master); exit(0); } Index: projects/clang-trunk/usr.sbin/yppush/yppush_main.c =================================================================== --- projects/clang-trunk/usr.sbin/yppush/yppush_main.c (revision 283870) +++ projects/clang-trunk/usr.sbin/yppush/yppush_main.c (revision 283871) @@ -1,620 +1,620 @@ /* * Copyright (c) 1995 * Bill Paul . All rights reserved. * * Redistribution and use in source and binary forms, with or without * modification, are permitted provided that the following conditions * are met: * 1. Redistributions of source code must retain the above copyright * notice, this list of conditions and the following disclaimer. * 2. Redistributions in binary form must reproduce the above copyright * notice, this list of conditions and the following disclaimer in the * documentation and/or other materials provided with the distribution. * 3. All advertising materials mentioning features or use of this software * must display the following acknowledgement: * This product includes software developed by Bill Paul. * 4. Neither the name of the author nor the names of any co-contributors * may be used to endorse or promote products derived from this software * without specific prior written permission. * * THIS SOFTWARE IS PROVIDED BY Bill Paul AND CONTRIBUTORS ``AS IS'' AND * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE * ARE DISCLAIMED. IN NO EVENT SHALL Bill Paul OR CONTRIBUTORS BE LIABLE * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF * SUCH DAMAGE. */ #include __FBSDID("$FreeBSD$"); #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include "ypxfr_extern.h" #include "yppush_extern.h" char *progname = "yppush"; int debug = 1; int _rpcpmstart = 0; char *yp_dir = _PATH_YP; char *yppush_mapname = NULL; /* Map to transfer. */ char *yppush_domain = NULL; /* Domain in which map resides. */ char *yppush_master = NULL; /* Master NIS server for said domain. */ int skip_master = 0; /* Do not attempt to push map to master. */ int verbose = 0; /* Toggle verbose mode. */ unsigned long yppush_transid = 0; int yppush_timeout = 80; /* Default timeout. */ int yppush_jobs = 1; /* Number of allowed concurrent jobs. */ int yppush_running_jobs = 0; /* Number of currently running jobs. */ /* Structure for holding information about a running job. */ struct jobs { unsigned long tid; int port; ypxfrstat stat; unsigned long prognum; char *server; char *map; int polled; struct jobs *next; }; struct jobs *yppush_joblist; /* Linked list of running jobs. */ static int yppush_svc_run(int); /* * Local error messages. */ static const char * yppusherr_string(int err) { switch (err) { case YPPUSH_TIMEDOUT: return("transfer or callback timed out"); case YPPUSH_YPSERV: return("failed to contact ypserv"); case YPPUSH_NOHOST: return("no such host"); case YPPUSH_PMAP: return("portmapper failure"); default: return("unknown error code"); } } /* * Report state of a job. */ static int yppush_show_status(ypxfrstat status, unsigned long tid) { struct jobs *job; job = yppush_joblist; while (job != NULL) { if (job->tid == tid) break; job = job->next; } if (job == NULL) { yp_error("warning: received callback with invalid transaction ID: %lu", tid); return (0); } if (job->polled) { yp_error("warning: received callback with duplicate transaction ID: %lu", tid); return (0); } if (verbose > 1) { yp_error("checking return status: transaction ID: %lu", job->tid); } - if (status != YPPUSH_SUCC || verbose) { + if (status != YPXFR_SUCC || verbose) { yp_error("transfer of map %s to server %s %s", - job->map, job->server, status == YPPUSH_SUCC ? + job->map, job->server, status == YPXFR_SUCC ? "succeeded" : "failed"); - yp_error("status returned by ypxfr: %s", status > YPPUSH_AGE ? + yp_error("status returned by ypxfr: %s", status > YPXFR_AGE ? yppusherr_string(status) : ypxfrerr_string(status)); } job->polled = 1; svc_unregister(job->prognum, 1); yppush_running_jobs--; return(0); } /* Exit routine. */ static void yppush_exit(int now) { struct jobs *jptr; int still_pending = 1; /* Let all the information trickle in. */ while (!now && still_pending) { jptr = yppush_joblist; still_pending = 0; while (jptr) { if (jptr->polled == 0) { still_pending++; if (verbose > 1) yp_error("%s has not responded", jptr->server); } else { if (verbose > 1) yp_error("%s has responded", jptr->server); } jptr = jptr->next; } if (still_pending) { if (verbose > 1) yp_error("%d transfer%sstill pending", still_pending, still_pending > 1 ? "s " : " "); if (yppush_svc_run (YPPUSH_RESPONSE_TIMEOUT) == 0) { yp_error("timed out"); now = 1; } } else { if (verbose) yp_error("all transfers complete"); break; } } /* All stats collected and reported -- kill all the stragglers. */ jptr = yppush_joblist; while (jptr) { if (!jptr->polled) yp_error("warning: exiting with transfer \ to %s (transid = %lu) still pending", jptr->server, jptr->tid); svc_unregister(jptr->prognum, 1); jptr = jptr->next; } exit(0); } /* * Handler for 'normal' signals. */ static void handler(int sig) { yppush_exit (1); return; } /* * Dispatch loop for callback RPC services. * Return value: * -1 error * 0 timeout * >0 request serviced */ static int yppush_svc_run(int timeout_secs) { int rc; fd_set readfds; struct timeval timeout; timeout.tv_usec = 0; timeout.tv_sec = timeout_secs; retry: readfds = svc_fdset; rc = select(svc_maxfd + 1, &readfds, NULL, NULL, &timeout); switch (rc) { case -1: if (errno == EINTR) goto retry; yp_error("select failed: %s", strerror(errno)); break; case 0: yp_error("select() timed out"); break; default: svc_getreqset(&readfds); break; } return rc; } /* * RPC service routines for callbacks. */ void * yppushproc_null_1_svc(void *argp, struct svc_req *rqstp) { static char * result; /* Do nothing -- RPC conventions call for all a null proc. */ return((void *) &result); } void * yppushproc_xfrresp_1_svc(yppushresp_xfr *argp, struct svc_req *rqstp) { static char * result; yppush_show_status(argp->status, argp->transid); return((void *) &result); } /* * Transmit a YPPROC_XFR request to ypserv. */ static int yppush_send_xfr(struct jobs *job) { ypreq_xfr req; /* ypresp_xfr *resp; */ DBT key, data; CLIENT *clnt; struct rpc_err err; struct timeval timeout; timeout.tv_usec = 0; timeout.tv_sec = 0; /* * The ypreq_xfr structure has a member of type map_parms, * which seems to require the order number of the map. * It isn't actually used at the other end (at least the * FreeBSD ypserv doesn't use it) but we fill it in here * for the sake of completeness. */ key.data = "YP_LAST_MODIFIED"; key.size = sizeof ("YP_LAST_MODIFIED") - 1; if (yp_get_record(yppush_domain, yppush_mapname, &key, &data, 1) != YP_TRUE) { yp_error("failed to read order number from %s: %s: %s", yppush_mapname, yperr_string(yp_errno), strerror(errno)); return(1); } /* Fill in the request arguments */ req.map_parms.ordernum = atoi(data.data); req.map_parms.domain = yppush_domain; req.map_parms.peer = yppush_master; req.map_parms.map = job->map; req.transid = job->tid; req.prog = job->prognum; req.port = job->port; /* Get a handle to the remote ypserv. */ if ((clnt = clnt_create(job->server, YPPROG, YPVERS, "udp")) == NULL) { yp_error("%s: %s",job->server,clnt_spcreateerror("couldn't \ create udp handle to NIS server")); switch (rpc_createerr.cf_stat) { case RPC_UNKNOWNHOST: job->stat = YPPUSH_NOHOST; break; case RPC_PMAPFAILURE: job->stat = YPPUSH_PMAP; break; default: job->stat = YPPUSH_RPC; break; } return(1); } /* * Reduce timeout to nothing since we may not * get a response from ypserv and we don't want to block. */ if (clnt_control(clnt, CLSET_TIMEOUT, (char *)&timeout) == FALSE) yp_error("failed to set timeout on ypproc_xfr call"); /* Invoke the ypproc_xfr service. */ if (ypproc_xfr_2(&req, clnt) == NULL) { clnt_geterr(clnt, &err); if (err.re_status != RPC_SUCCESS && err.re_status != RPC_TIMEDOUT) { yp_error("%s: %s", job->server, clnt_sperror(clnt, "yp_xfr failed")); job->stat = YPPUSH_YPSERV; clnt_destroy(clnt); return(1); } } clnt_destroy(clnt); return(0); } /* * Main driver function. Register the callback service, add the transfer * request to the internal list, send the YPPROC_XFR request to ypserv * do other magic things. */ -int +static int yp_push(char *server, char *map, unsigned long tid) { unsigned long prognum; int sock = RPC_ANYSOCK; SVCXPRT *xprt; struct jobs *job; /* Register the job in our linked list of jobs. */ /* First allocate job structure */ if ((job = (struct jobs *)malloc(sizeof (struct jobs))) == NULL) { yp_error("malloc failed"); yppush_exit (1); } /* * Register the callback service on the first free transient * program number. */ xprt = svcudp_create(sock); for (prognum = 0x40000000; prognum < 0x5FFFFFFF; prognum++) { if (svc_register(xprt, prognum, 1, yppush_xfrrespprog_1, IPPROTO_UDP) == TRUE) break; } if (prognum == 0x5FFFFFFF) { yp_error ("can't register yppush_xfrrespprog_1"); yppush_exit (1); } /* Initialize the info for this job. */ job->stat = 0; job->tid = tid; job->port = xprt->xp_port; job->server = strdup(server); job->map = strdup(map); job->prognum = prognum; job->polled = 0; job->next = yppush_joblist; yppush_joblist = job; if (verbose) { yp_error("initiating transfer: %s -> %s (transid = %lu)", yppush_mapname, server, tid); } /* * Send the XFR request to ypserv. We don't have to wait for * a response here since we handle them asynchronously. */ if (yppush_send_xfr(job)){ /* Transfer request blew up. */ yppush_show_status(job->stat ? job->stat : YPPUSH_YPSERV,job->tid); } else { if (verbose > 1) yp_error("%s has been called", server); } return(0); } /* * Called for each entry in the ypservers map from yp_get_map(), which * is our private yp_all() routine. */ -int +static int yppush_foreach(int status, char *key, int keylen, char *val, int vallen, char *data) { char server[YPMAXRECORD + 2]; if (status != YP_TRUE) return (status); snprintf(server, sizeof(server), "%.*s", vallen, val); if (skip_master && strcasecmp(server, yppush_master) == 0) return (0); /* * Restrict the number of concurrent jobs: if yppush_jobs number * of jobs have already been dispatched and are still pending, * wait for one of them to finish so we can reuse its slot. */ while (yppush_running_jobs >= yppush_jobs && (yppush_svc_run (yppush_timeout) > 0)) ; /* Cleared for takeoff: set everything in motion. */ if (yp_push(server, yppush_mapname, yppush_transid)) return(yp_errno); /* Bump the job counter and transaction ID. */ yppush_running_jobs++; yppush_transid++; return (0); } static void usage() { fprintf (stderr, "%s\n%s\n", "usage: yppush [-d domain] [-t timeout] [-j #parallel jobs] [-h host]", " [-p path] mapname"); exit(1); } /* * Entry point. (About time!) */ int main(int argc, char *argv[]) { int ch; DBT key, data; char myname[MAXHOSTNAMELEN]; struct hostlist { char *name; struct hostlist *next; }; struct hostlist *yppush_hostlist = NULL; struct hostlist *tmp; while ((ch = getopt(argc, argv, "d:j:p:h:t:v")) != -1) { switch (ch) { case 'd': yppush_domain = optarg; break; case 'j': yppush_jobs = atoi(optarg); if (yppush_jobs <= 0) yppush_jobs = 1; break; case 'p': yp_dir = optarg; break; case 'h': /* we can handle multiple hosts */ if ((tmp = (struct hostlist *)malloc(sizeof(struct hostlist))) == NULL) { yp_error("malloc failed"); yppush_exit(1); } tmp->name = strdup(optarg); tmp->next = yppush_hostlist; yppush_hostlist = tmp; break; case 't': yppush_timeout = atoi(optarg); break; case 'v': verbose++; break; default: usage(); break; } } argc -= optind; argv += optind; yppush_mapname = argv[0]; if (yppush_mapname == NULL) { /* "No guts, no glory." */ usage(); } /* * If no domain was specified, try to find the default * domain. If we can't find that, we're doomed and must bail. */ if (yppush_domain == NULL) { char *yppush_check_domain; if (!yp_get_default_domain(&yppush_check_domain) && !_yp_check(&yppush_check_domain)) { yp_error("no domain specified and NIS not running"); usage(); } else yp_get_default_domain(&yppush_domain); } /* Check to see that we are the master for this map. */ if (gethostname ((char *)&myname, sizeof(myname))) { yp_error("failed to get name of local host: %s", strerror(errno)); yppush_exit(1); } key.data = "YP_MASTER_NAME"; key.size = sizeof("YP_MASTER_NAME") - 1; if (yp_get_record(yppush_domain, yppush_mapname, &key, &data, 1) != YP_TRUE) { yp_error("couldn't open %s map: %s", yppush_mapname, strerror(errno)); yppush_exit(1); } if (strncasecmp(myname, data.data, data.size) == 0) { /* I am master server, and no explicit host list was specified: do not push map to myself -- this will fail with YPPUSH_AGE anyway. */ if (yppush_hostlist == NULL) skip_master = 1; } else { yp_error("warning: this host is not the master for %s", yppush_mapname); #ifdef NITPICKY yppush_exit(1); #endif } yppush_master = malloc(data.size + 1); strncpy(yppush_master, data.data, data.size); yppush_master[data.size] = '\0'; /* Install some handy handlers. */ signal(SIGTERM, handler); signal(SIGINT, handler); /* set initial transaction ID */ yppush_transid = time((time_t *)NULL); if (yppush_hostlist) { /* * Host list was specified on the command line: * kick off the transfers by hand. */ tmp = yppush_hostlist; while (tmp) { yppush_foreach(YP_TRUE, NULL, 0, tmp->name, strlen(tmp->name), NULL); tmp = tmp->next; } } else { /* * Do a yp_all() on the ypservers map and initiate a ypxfr * for each one. */ ypxfr_get_map("ypservers", yppush_domain, "localhost", yppush_foreach); } if (verbose > 1) yp_error("all jobs dispatched"); /* All done -- normal exit. */ yppush_exit(0); /* Just in case. */ exit(0); } Index: projects/clang-trunk =================================================================== --- projects/clang-trunk (revision 283870) +++ projects/clang-trunk (revision 283871) Property changes on: projects/clang-trunk ___________________________________________________________________ Modified: svn:mergeinfo ## -0,0 +0,1 ## Merged /head:r283771-283870